{ "brief_id": "SPR-2026-4728", "generated_at": "2026-09-24", "domains": [ "Plant and animal studies", "Plant Pathogens and Fungal Diseases" ], "original_hypothesis": "The antagonistic potential of plant endophytes against pathogens is phylogenetically conserved and predictable from fungal phylogenies, enabling the engineering of synthetic endophytic communities for crop protection.", "grounding": { "novelty_assessment": { "score": 0.72, "closest_existing_work": [ { "paper_id": "410f2958fbf4a6dd2cb18083dc74aaf1551a2611", "title": "A conceptual framework for the phylogenetically constrained assembly of microbial communities", "doi": "10.1186/s40168-019-0754-y", "year": 2019, "similarity": "high", "key_difference": "This paper establishes that microbial community assembly is phylogenetically constrained and that traits show phylogenetic signal, but it does not specifically address endophytic fungal antagonism against plant pathogens, nor does it propose predicting biocontrol efficacy from phylogeny or engineering synthetic communities for crop protection.", "authors": [ "Daniel Aguirre de Cárcer" ] }, { "paper_id": "1787d990bb49ff9af27db169c637c2966d1c74da", "title": "Engineering the plant microbiome: synthetic community approaches to enhance crop protection", "doi": "10.3389/fpls.2025.1705289", "year": 2026, "similarity": "high", "key_difference": "This paper reviews SynCom design for crop protection using ecological principles and computational tools, but it does not propose phylogenetic conservation of antagonistic traits as a predictive framework for selecting untested endophytes, nor does it link agricultural intensification to erosion of phylogenetically conserved biocontrol functions.", "authors": [ "Ketankumar J. Panchal", "Ankit P. Sudhir", "A. Prajapati" ] }, { "paper_id": "dce142a6eb9141c8d6151c116a2b8d06d723bcda", "title": "Inhibition of plant pathogenic fungi by endophytic Trichoderma spp. through mycoparasitism and volatile organic compounds.", "doi": "10.1016/j.micres.2020.126595", "year": 2020, "similarity": "moderate", "key_difference": "This paper demonstrates antagonistic mechanisms within a single genus (Trichoderma), providing a case study consistent with phylogenetic conservation, but it does not test phylogenetic signal across the fungal tree of life or propose a predictive framework.", "authors": [ "P. Rajani", "C. Rajasekaran", "M. M. Vasanthakumari", "Shannon B. Olsson", "G. Ravikanth", "R. Shaanker" ] }, { "paper_id": "b9faa52207308d98f17b4cacbccb368cd7a52648", "title": "Mycoparasitism of Endophytic Fungi Isolated From Reed on Soilborne Phytopathogenic Fungi and Production of Cell Wall-Degrading Enzymes In Vitro", "doi": "10.1007/s00284-009-9477-9", "year": 2009, "similarity": "moderate", "key_difference": "This paper documents mycoparasitism by endophytic fungi from reed against soilborne pathogens, but it does not examine phylogenetic relationships among the endophytes or test whether antagonistic traits are phylogenetically conserved.", "authors": [ "Ronghua Cao", "Xiao-Guang Liu", "K. Gao", "K. Mendgen", "Z. Kang", "Jianfeng Gao", "Yang Dai", "Xue Wang" ] } ], "verdict": "novel" }, "evidence_base": [ { "paper_id": "dce142a6eb9141c8d6151c116a2b8d06d723bcda", "title": "Inhibition of plant pathogenic fungi by endophytic Trichoderma spp. through mycoparasitism and volatile organic compounds.", "doi": "10.1016/j.micres.2020.126595", "year": 2020, "citation_count": 171, "support_type": "indirect", "relevance": "Demonstrates that multiple endophytic Trichoderma species share antagonistic mechanisms (mycoparasitism, VOCs) against plant pathogens, consistent with phylogenetic conservation within a genus.", "key_finding": "Four endophytic Trichoderma spp. inhibit pathogenic fungi through mycoparasitism and volatile organic compounds.", "authors": [ "P. Rajani", "C. Rajasekaran", "M. M. Vasanthakumari", "Shannon B. Olsson", "G. Ravikanth", "R. Shaanker" ] }, { "paper_id": "2ab913c5595abecd16a54d10977eba44af6e3768", "title": "Endophytic Trichoderma gamsii YIM PH30019: a promising biocontrol agent with hyperosmolar, mycoparasitism, and antagonistic activities of induced volatile organic compounds on root-rot pathogenic fungi of Panax notoginseng", "doi": "10.1016/j.jgr.2015.09.006", "year": 2015, "citation_count": 156, "support_type": "indirect", "relevance": "Shows that an endophytic Trichoderma species exhibits multiple antagonistic activities (mycoparasitism, VOCs) against root-rot pathogens, supporting the idea that antagonistic traits are shared among closely related endophytes.", "key_finding": "Trichoderma gamsii YIM PH30019 has biocontrol potential against notoginseng phytodiseases via mycoparasitism and VOCs.", "authors": [ "Jin-Lian Chen", "Shizhong Sun", "Cui-Ping Miao", "Kai Wu", "You-Wei Chen", "Li-Hua Xu", "Hui-Lin Guan", "Li-Xing Zhao" ] }, { "paper_id": "b9faa52207308d98f17b4cacbccb368cd7a52648", "title": "Mycoparasitism of Endophytic Fungi Isolated From Reed on Soilborne Phytopathogenic Fungi and Production of Cell Wall-Degrading Enzymes In Vitro", "doi": "10.1007/s00284-009-9477-9", "year": 2009, "citation_count": 67, "support_type": "indirect", "relevance": "Documents mycoparasitism as a mechanism of antagonism by endophytic fungi, one of the mechanisms proposed to be phylogenetically conserved.", "key_finding": "Three endophytic fungi from reed inhibit soilborne pathogens by coiling around hyphae and degrading hyphal cytoplasm.", "authors": [ "Ronghua Cao", "Xiao-Guang Liu", "K. Gao", "K. Mendgen", "Z. Kang", "Jianfeng Gao", "Yang Dai", "Xue Wang" ] }, { "paper_id": "410f2958fbf4a6dd2cb18083dc74aaf1551a2611", "title": "A conceptual framework for the phylogenetically constrained assembly of microbial communities", "doi": "10.1186/s40168-019-0754-y", "year": 2019, "citation_count": 40, "support_type": "direct", "relevance": "Provides the theoretical foundation that microbial traits and ecological coherence are phylogenetically conserved, directly supporting the hypothesis that antagonistic traits in endophytic fungi are phylogenetically conserved.", "key_finding": "Microbial community assembly is phylogenetically constrained, and traits show significant phylogenetic signal.", "authors": [ "Daniel Aguirre de Cárcer" ] }, { "paper_id": "1787d990bb49ff9af27db169c637c2966d1c74da", "title": "Engineering the plant microbiome: synthetic community approaches to enhance crop protection", "doi": "10.3389/fpls.2025.1705289", "year": 2026, "citation_count": 29, "support_type": "indirect", "relevance": "Discusses SynCom design for crop protection, supporting the feasibility of engineering synthetic endophytic communities, but does not address phylogenetic prediction.", "key_finding": "SynComs can be rationally designed using ecological principles and computational tools for crop protection.", "authors": [ "Ketankumar J. Panchal", "Ankit P. Sudhir", "A. Prajapati" ] }, { "paper_id": "d5d3b7fb91d6fa717fcfdf9d9faa4fa268a3b04d", "title": "Epigenetic Activation of Silent Biosynthetic Gene Clusters in Endophytic Fungi Using Small Molecular Modifiers", "doi": "10.3389/fmicb.2022.815008", "year": 2022, "citation_count": 48, "support_type": "indirect", "relevance": "Highlights that endophytic fungi harbor biosynthetic gene clusters for secondary metabolites, which could underlie phylogenetic conservation of antagonistic traits.", "key_finding": "Fungal endophytes contain silent biosynthetic gene clusters that can be activated to produce secondary metabolites.", "authors": [ "Lynise C. Pillay", "Lucpah Nekati", "Phuti J. Makhwitine", "S. Ndlovu" ] }, { "paper_id": "c8f506a38050613eb11a2c21112e1b491a614e01", "title": "Metabolomic-guided discovery of cyclic nonribosomal peptides from Xylaria ellisii sp. nov., a leaf and stem endophyte of Vaccinium angustifolium", "doi": "10.1038/s41598-020-61088-x", "year": 2020, "citation_count": 30, "support_type": "indirect", "relevance": "Shows that an endophytic Xylaria species produces novel antimicrobial compounds, supporting the link between biosynthetic gene clusters and antagonistic potential.", "key_finding": "Xylaria ellisii produces eight new cyclic nonribosomal peptides with potential bioactivity.", "authors": [ "A. Ibrahim", "J. Tanney", "Fan Fei", "K. Seifert", "G. Cutler", "A. Capretta", "J. Miller", "Mark W. Sumarah" ] }, { "paper_id": "17185eecc446c02c8410f374071ff093c132cf28", "title": "Biodiversity–production feedback effects lead to intensification traps in agricultural landscapes", "doi": "10.1038/s41559-024-02349-0", "year": 2024, "citation_count": 47, "support_type": "analogous", "relevance": "Provides evidence that agricultural intensification leads to biodiversity loss and production declines, analogous to the proposed erosion of endophyte-mediated biocontrol services.", "key_finding": "Intensive agriculture can trigger intensification traps due to biodiversity loss feedback on crop yields.", "authors": [ "Alfred Burian", "C. Kremen", "James Shyan-Tau Wu", "Michael Beckmann", "M. Bulling", "L. Garibaldi", "Tamás Krisztin", "Z. Mehrabi", "N. Ramankutty", "R. Seppelt" ] } ], "counter_evidence": [ { "paper_id": "71538bb2624169608f10dae937d052c0c2a69e42", "title": "A bacterial endophyte exploits chemotropism of a fungal pathogen for plant colonization", "doi": "10.1038/s41467-020-18994-5", "year": 2020, "finding": "This study shows that a bacterial endophyte (Rahnella aquatilis) antagonizes a fungal pathogen through pH-mediated chemotaxis and exploitation of hyphae, a mechanism not necessarily phylogenetically conserved among fungi. It highlights that antagonistic mechanisms can be highly specific and not predictable from fungal phylogeny alone.", "severity": "minor", "authors": [ "D. Palmieri", "Stefania Vitale", "G. Lima", "A. Di Pietro", "D. Turrà" ] }, { "paper_id": "dc2f619f9d92aec4d68a790f6244172a78942868", "title": "Towards unlocking the biocontrol potential of Pichia kudriavzevii for plant fungal diseases: in vitro and in vivo assessments with candidate secreted protein prediction", "doi": "10.1186/s12866-023-03047-w", "year": 2023, "finding": "This paper identifies biocontrol potential in Pichia kudriavzevii, a yeast, and predicts secreted proteins as candidate effectors. It suggests that antagonistic traits may be mediated by secreted proteins that could be species-specific rather than broadly conserved across fungal phylogeny.", "severity": "minor", "authors": [ "Bassma Mahmoud Elkhairy", "Nabil Mohamed Salama", "Abdalrahman Mohammad Desouki", "A. Abdelrazek", "K. A. Soliman", "Samir Abdelaziz Ibrahim", "H. Khalil" ] } ], "gap_manifest_update": { "closed_gaps": [], "new_gaps": [ "Lack of empirical studies testing phylogenetic signal (e.g., Pagel's λ, Blomberg's K) for antagonistic traits across a broad phylogeny of endophytic fungi.", "No standardized framework for linking biosynthetic gene clusters to antagonistic phenotypes in endophytic fungi across diverse taxa.", "Limited understanding of how agricultural intensification quantitatively affects endophytic fungal community composition and functional antagonistic potential across multiple crops and regions." ], "data_available": [ "Qualitative and semi-quantitative antagonism data from in vitro plate assays (e.g., inhibition zones, mycoparasitism observations) in papers such as dce142a6eb9141c8d6151c116a2b8d06d723bcda, 2ab913c5595abecd16a54d10977eba44af6e3768, and b9faa52207308d98f17b4cacbccb368cd7a52648.", "Phylogenetic frameworks for microbial community assembly (410f2958fbf4a6dd2cb18083dc74aaf1551a2611).", "Biosynthetic gene cluster data for some endophytic fungi (d5d3b7fb91d6fa717fcfdf9d9faa4fa268a3b04d, c8f506a38050613eb11a2c21112e1b491a614e01)." ] }, "search_queries": [ { "query": "phylogenetic conservation antagonistic endophytes", "type": "novelty", "rationale": "Direct search for the core hypothesis on phylogenetic conservation of antagonistic traits in endophytes" }, { "query": "predicting biocontrol agents fungal phylogeny", "type": "novelty", "rationale": "Tests whether the predictive framework using fungal phylogenies for biocontrol agent selection has been proposed" }, { "query": "synthetic endophytic communities crop protection", "type": "novelty", "rationale": "Checks for existing work on engineering synthetic endophytic communities for crop protection" }, { "query": "phylogenetically guided endophyte discovery", "type": "novelty", "rationale": "Directly searches for the concept of phylogenetically guided discovery of endophytes as biocontrol agents" }, { "query": "endophyte antibiosis pathogen inhibition", "type": "evidence", "rationale": "Finds evidence for antibiosis as a mechanism of pathogen antagonism by endophytes" }, { "query": "mycoparasitism endophytic fungi", "type": "evidence", "rationale": "Finds evidence for mycoparasitism as an antagonistic mechanism of endophytic fungi" }, { "query": "induced systemic resistance endophytes", "type": "evidence", "rationale": "Finds evidence for induced systemic resistance in host plants triggered by endophytes" }, { "query": "biosynthetic gene clusters endophytic fungi", "type": "evidence", "rationale": "Finds evidence for conserved biosynthetic gene clusters in endophytic fungi related to antimicrobial compounds" }, { "query": "phylogenetic signal microbial traits", "type": "cross_domain", "rationale": "Borrows methods from microbial ecology to test phylogenetic signal in traits, applicable to endophyte antagonism" }, { "query": "agricultural intensification biodiversity loss biocontrol", "type": "cross_domain", "rationale": "Links agricultural intensification, biodiversity loss, and disruption of biocontrol services" }, { "query": "microbiome engineering plant health", "type": "cross_domain", "rationale": "Draws from microbiome engineering approaches in plant health for synthetic community design" } ] }, "sharpened": { "title": "Phylogenetic signal in fungal endophyte antagonism predicts synthetic community efficacy against plant pathogens", "formal_statement": "If antagonistic traits of fungal endophytes against plant pathogens exhibit a significant phylogenetic signal (Pagel's λ ≥ 0.3) across the fungal tree of life, then phylogenetically guided selection of untested endophytes will predict their in vitro and in planta antagonistic efficacy better than random selection, because shared biosynthetic gene clusters and conserved ecological strategies underlie trait inheritance.", "independent_variables": [ { "name": "Phylogenetic distance between donor and recipient endophyte species", "type": "continuous", "range": "0.01-1.5", "unit": "substitutions per site (branch length)" }, { "name": "Endophyte species identity (phylogenetic position)", "type": "categorical", "range": "≥100 species across Ascomycota and Basidiomycota", "unit": "taxon ID" }, { "name": "Antagonistic mechanism class", "type": "categorical", "range": "antibiosis, mycoparasitism, VOC production, induced systemic resistance", "unit": "mechanism category" }, { "name": "Pathogen focal species", "type": "categorical", "range": "Fusarium graminearum, Botrytis cinerea, Rhizoctonia solani, Magnaporthe oryzae", "unit": "pathogen ID" }, { "name": "Agricultural intensification index", "type": "ordinal", "range": "low (extensive), medium (integrated), high (conventional)", "unit": "categorical score 1-3" } ], "dependent_variables": [ { "name": "In vitro antagonism score (dual-culture inhibition)", "type": "continuous", "expected_direction": "increase", "unit": "percentage of pathogen radial growth inhibition (%)" }, { "name": "Mycoparasitism incidence", "type": "continuous", "expected_direction": "increase", "unit": "percentage of hyphal contact sites with coiling/lysis (%)" }, { "name": "VOC-mediated inhibition", "type": "continuous", "expected_direction": "increase", "unit": "percentage of pathogen biomass reduction in split-plate assay (%)" }, { "name": "Antibiosis zone of clearance", "type": "continuous", "expected_direction": "increase", "unit": "mm of inhibition halo" }, { "name": "In planta disease severity", "type": "continuous", "expected_direction": "decrease", "unit": "lesion area (mm²) or disease index (0-100)" }, { "name": "Pathogen load (qPCR)", "type": "continuous", "expected_direction": "decrease", "unit": "pathogen DNA copies per ng plant DNA" }, { "name": "Phylogenetic signal (Pagel's λ)", "type": "continuous", "expected_direction": "non-monotonic", "unit": "dimensionless (0-1)" }, { "name": "SynCom protection efficacy", "type": "continuous", "expected_direction": "increase", "unit": "percentage reduction in disease relative to untreated control (%)" } ], "proposed_mechanism": { "causal_chain": [ "Step 1: Endophytic fungi possess biosynthetic gene clusters (BGCs) and effector genes encoding antagonistic traits (antibiotics, cell wall-degrading enzymes, VOC synthases).", "Step 2: These genes are vertically inherited and subject to purifying selection, leading to phylogenetic conservation of antagonistic phenotypes within clades.", "Step 3: The strength and specificity of antagonism against a given pathogen depend on the phylogenetic distance between endophyte and pathogen, as well as the endophyte's phylogenetic position.", "Step 4: Phylogenetic signal (Pagel's λ) quantifies the degree to which antagonistic trait variation is explained by shared ancestry across the fungal tree of life.", "Step 5: A predictive model incorporating phylogeny, BGC content, and mechanism class can forecast the antagonistic efficacy of untested endophytes against focal pathogens.", "Step 6: Synthetic communities assembled from phylogenetically guided predictions will suppress disease more effectively than random communities of equivalent diversity.", "Step 7: Agricultural intensification reduces endophyte diversity and disrupts phylogenetic structure, eroding natural biocontrol services." ], "key_assumptions": [ "Antagonistic traits are measurable in vitro and correlate with in planta efficacy.", "The fungal phylogeny used is accurate and resolves relationships at the species level.", "Horizontal gene transfer (HGT) of BGCs does not overwhelm vertical inheritance signal (HGT rate < 10% of trait variance).", "Environmental conditions (pH, water activity, light) are standardized and do not differentially affect endophyte growth.", "Pathogen strains used are representative of field populations." ], "known_unknowns": [ "The exact rate of HGT for antagonistic BGCs among endophytic fungi.", "Whether phylogenetic signal is consistent across different pathogen taxa.", "The extent to which epigenetic regulation silences BGCs in vitro.", "The long-term stability of SynComs in field conditions.", "The impact of agricultural intensification on phylogenetic diversity of endophytes." ] }, "falsifiable_predictions": [ { "prediction": "Pagel's λ for in vitro antagonism score against Fusarium graminearum is significantly greater than 0 and less than 1.", "quantitative_bound": "λ = 0.3-0.7 (95% CI), p < 0.01", "measurement_method": "Phylogenetic signal analysis using Pagel's λ on a time-calibrated tree of ≥100 species with antagonism scores from dual-culture assays.", "null_hypothesis": "H0: λ = 0 (no phylogenetic signal; trait evolves independently of phylogeny).", "statistical_test": "Likelihood ratio test comparing λ = 0 vs. λ estimated, alpha = 0.05, with 1000 simulations under Brownian motion." }, { "prediction": "Phylogenetically guided prediction of antagonism (based on trait values of close relatives) explains more variance in observed antagonism than random prediction.", "quantitative_bound": "R² ≥ 0.4 for phylogenetic prediction vs. R² ≤ 0.1 for random, difference significant at p < 0.001", "measurement_method": "Cross-validated phylogenetic imputation (e.g., using phylolm or RPANDA) on held-out species; compare predicted vs. observed antagonism scores.", "null_hypothesis": "H0: R²_phylogenetic = R²_random (no difference in predictive power).", "statistical_test": "Paired t-test or Wilcoxon signed-rank test on cross-validated R² values, alpha = 0.05." }, { "prediction": "SynComs designed using phylogenetic predictions reduce disease severity by at least 30% more than random SynComs of equivalent species richness.", "quantitative_bound": "Mean disease reduction: predicted SynCom = 50-70%, random SynCom = 20-40%, difference ≥ 30 percentage points (95% CI: 15-45), p < 0.01", "measurement_method": "In planta greenhouse trials with focal pathogen (e.g., Fusarium graminearum on wheat), disease index measured at 14 days post-inoculation; compare predicted vs. random SynComs.", "null_hypothesis": "H0: no difference in disease reduction between predicted and random SynComs.", "statistical_test": "Linear mixed-effects model with SynCom type as fixed effect and block as random effect, ANOVA, alpha = 0.05." }, { "prediction": "The strength of phylogenetic signal (λ) is clade-specific and higher for antibiosis than for mycoparasitism.", "quantitative_bound": "λ_antibiosis = 0.5-0.8, λ_mycoparasitism = 0.1-0.4, difference significant at p < 0.05", "measurement_method": "Separate Pagel's λ estimates for each mechanism class using trait data from ≥100 species; compare λ values with likelihood ratio tests.", "null_hypothesis": "H0: λ_antibiosis = λ_mycoparasitism (no difference in phylogenetic signal between mechanisms).", "statistical_test": "Likelihood ratio test comparing models with equal vs. different λ for each mechanism, alpha = 0.05." }, { "prediction": "Agricultural intensification index is negatively correlated with endophyte phylogenetic diversity and biocontrol potential.", "quantitative_bound": "Spearman's ρ ≤ -0.5 between intensification index and phylogenetic diversity (Faith's PD) or mean antagonism score, p < 0.01", "measurement_method": "Field survey of endophytes from crops across intensification gradients; measure phylogenetic diversity and in vitro antagonism against focal pathogens.", "null_hypothesis": "H0: ρ = 0 (no correlation between intensification and endophyte phylogenetic diversity or antagonism).", "statistical_test": "Spearman's rank correlation, alpha = 0.05, with correction for multiple comparisons." } ], "boundary_conditions": [ { "condition": "In vitro assays must be performed at pH 6.5-7.0, water activity 0.98-1.00, and 22-25°C under 12h light/dark cycles.", "justification": "These conditions standardize growth and antagonism expression; deviations may alter BGC expression and volatile production." }, { "condition": "Phylogenetic analysis requires a well-resolved tree with ≥100 species and <10% missing data.", "justification": "Insufficient taxon sampling or unresolved nodes reduce power to detect phylogenetic signal." }, { "condition": "HGT rate for antagonistic BGCs must be below 10% of trait variance.", "justification": "High HGT can obscure vertical inheritance signal, invalidating phylogenetic prediction." }, { "condition": "In planta trials must use a single pathogen strain and a susceptible host genotype.", "justification": "Host resistance and pathogen variability can confound SynCom efficacy." }, { "condition": "SynComs must be assembled with species that are culturable and stable in co-culture.", "justification": "Non-culturable or incompatible species cannot be used in synthetic communities." } ], "theoretical_framework": "Phylogenetic niche conservatism and community assembly theory" }, "protocol": { "protocol_title": "Phylogenetic signal in fungal endophyte antagonism as predictor of synthetic community efficacy: a 3-phase validation protocol", "overall_timeline": "10-20 months", "overall_budget_estimate": "€18k-120k", "phases": [ { "phase_number": 1, "phase_name": "In Silico Validation", "objective": "Test whether antagonistic traits (in vitro inhibition, mycoparasitism, VOC, antibiosis) exhibit significant phylogenetic signal (Pagel's λ ≥ 0.3) across ≥100 endophyte species, and whether phylogeny-based imputation predicts antagonism better than random.", "methodology": "1) Assemble a curated trait database from published dual-culture, split-plate VOC, and mycoparasitism assays (literature mining + targeted re-extraction from 8 evidence-base papers and >200 additional studies via Web of Science/Scopus queries). 2) Build a time-calibrated species-level phylogeny of ≥100 Ascomycota/Basidiomycota endophytes using ITS+LSU+tef1+rpb2 sequences retrieved from NCBI GenBank; align with MAFFT v7, infer with IQ-TREE 2 (ModelFinder, 1000 UFBoot), date with treePL or BEAST 2 using published fungal calibration points. 3) Compute Pagel's λ, Blomberg's K, and phylogenetic half-life (phylolm) for each antagonism trait against each focal pathogen (F. graminearum, B. cinerea, R. solani, M. oryzae) using phytools, ape, and caper in R. 4) Run likelihood-ratio tests (λ=0 vs estimated) with 1000 Brownian-motion simulations. 5) Cross-validated phylogenetic imputation (phylolm, RPANDA, and a phylogenetic GLS with BGC-covariate) on 20% held-out species; compare R² vs random-shuffle baseline. 6) Test clade-specific λ for antibiosis vs mycoparasitism via multi-λ models (OUwie, phytools fitMultiBM). 7) Meta-analyze HGT rate for antagonistic BGCs from published fungal pangenome studies (antiSMASH 7 + BiG-SLiCE) to bound HGT contribution to trait variance.", "required_resources": { "equipment": [ "Workstation with ≥32 GB RAM (or institutional HPC allocation)" ], "software": [ "R 4.4+ with ape, phytools, caper, phylolm, RPANDA, OUwie, geiger", "IQ-TREE 2", "MAFFT v7", "BEAST 2 or treePL", "antiSMASH 7", "BiG-SLiCE", "Trimal", "FigTree" ], "datasets": [ "NCBI GenBank ITS/LSU/tef1/rpb2 sequences for ≥100 endophyte species", "Published antagonism trait tables (8 evidence-base papers + literature-mined)", "Fungal BGC catalogues (antiSMASH-DB, MIBiG 3.0)", "Time-calibrated fungal tree backbone (e.g., James et al. 2020, Naranjo-Ortiz & Gabaldón 2019)" ], "competences": [ "Phylogenetics (tree inference, dating, comparative methods)", "R scripting", "Literature mining / meta-analysis", "Fungal taxonomy" ], "estimated_cost": "€0-2000", "estimated_duration": "4-8 weeks" }, "expected_outputs": [ "Output 1: Curated trait matrix (≥100 species × 4 antagonism traits × 4 pathogens) with metadata", "Output 2: Time-calibrated species-level phylogeny (Newick + XML)", "Output 3: Pagel's λ, Blomberg's K, and 95% CI per trait × pathogen", "Output 4: Cross-validated R² for phylogenetic vs random imputation", "Output 5: HGT rate estimate and its upper bound on trait variance", "Output 6: Pre-registered Phase 2 design document" ], "success_criteria": [ { "metric": "Pagel's λ for in vitro antagonism vs F. graminearum", "threshold": "λ ≥ 0.3 with 95% CI excluding 0, p < 0.01 (LRT vs λ=0)", "measurement": "phytools::phylosig with 1000 simulations" }, { "metric": "Phylogenetic imputation R² vs random", "threshold": "R²_phylo ≥ 0.4 and R²_random ≤ 0.1, difference p < 0.001", "measurement": "phylolm cross-validation on 20% held-out species, Wilcoxon signed-rank" }, { "metric": "HGT contribution to trait variance", "threshold": "< 10% of total trait variance", "measurement": "antiSMASH/BiG-SLiCE BGC distribution vs phylogeny, Pagel's λ on BGC presence" }, { "metric": "Clade-specific λ difference (antibiosis vs mycoparasitism)", "threshold": "Δλ ≥ 0.2, p < 0.05", "measurement": "LRT comparing single-λ vs multi-λ models (OUwie)" } ], "go_nogo_decision": { "go_if": "λ ≥ 0.3 for at least 2 of 4 antagonism traits against ≥2 pathogens AND phylogenetic imputation R² ≥ 0.4 AND HGT < 10% variance", "nogo_if": "λ < 0.1 for all traits OR phylogenetic imputation R² ≤ 0.1 OR HGT > 30% variance (phylogenetic prediction invalid)", "pivot_if": "λ between 0.1-0.3 or R² 0.1-0.4: pivot to trait-based (BGC-content) prediction model without phylogenetic weighting, and re-scope Phase 2 to test BGC-based prediction instead" }, "risks": [ { "risk": "Insufficient published trait data for ≥100 species (missing data > 30%)", "probability": "high", "mitigation": "Expand to ≥150 candidate species; use phylogenetic imputation to fill gaps; supplement with targeted in-house assays in Phase 2" }, { "risk": "Poorly resolved or non-monophyletic species-level tree", "probability": "medium", "mitigation": "Use multi-locus (4+ genes) and coalescent-based inference (ASTRAL); collapse unresolved nodes; sensitivity analysis on alternative topologies" }, { "risk": "HGT of BGCs obscures vertical signal", "probability": "medium", "mitigation": "Explicitly model HGT via BGC-phylogeny discordance (cophylogenetic analysis with Jane 4 or eMPRess); if HGT > 10%, restrict to vertically inherited BGC families" }, { "risk": "Publication bias inflates trait values for well-studied genera (Trichoderma)", "probability": "high", "mitigation": "Weight by study sample size; run sensitivity analysis excluding Trichoderma-dominated clades; use meta-analytic random effects (metafor)" } ] }, { "phase_number": 2, "phase_name": "Minimal Experimental Validation", "objective": "Empirically confirm in vitro that (a) antagonistic traits are phylogenetically structured in a focal clade, and (b) phylogenetically guided selection predicts antagonism better than random, using the smallest tractable species set.", "methodology": "1) Select 24 endophyte species spanning the focal clade identified in Phase 1 (e.g., Hypocreales + Xylariales), including 12 species with published antagonism data and 12 untested close relatives (phylogenetic pairs at varying distances). 2) Obtain strains from public culture collections (CBS-KNAW, ATCC, DSMZ) or isolate from field-collected asymptomatic plant tissue (surface sterilization + PDA/MEA isolation). 3) Perform dual-culture assays against F. graminearum and B. cinerea on PDA at pH 6.8, aw 0.99, 23°C, 12h light/dark, 4 replicates; measure radial growth inhibition (%) at 7 days. 4) Split-plate VOC assay (I-plate) with pathogen biomass (dry weight) at 7 days. 5) Mycoparasitism incidence via light microscopy (coiling/lysis at ≥100 contact sites per pair). 6) Antibiosis halo (mm) on agar. 7) Re-estimate Pagel's λ on the 24-species subset and compute cross-validated R² for phylogenetic vs random prediction of the 12 untested species. 8) Confirm BGC presence via antiSMASH on draft genomes (Illumina 2×150, ~50× coverage) of the 24 strains.", "required_resources": { "equipment": [ "Biosafety level 2 mycology lab", "Laminar flow hood", "Incubators (12h light/dark, 23°C)", "Plate reader / image analysis station (ImageJ)", "Light microscope with camera", "Illumina MiSeq or NovaSeq (outsourced)", "qPCR instrument (optional for BGC expression)" ], "software": [ "ImageJ / Fiji for radial growth measurement", "R (phytools, phylolm)", "antiSMASH 7", "FastQC, SPAdes, BUSCO" ], "datasets": [ "24 endophyte strains (CBS/ATCC/DSMZ + field isolates)", "F. graminearum PH-1 and B. cinerea B05.10 reference strains", "Phase 1 trait matrix and phylogeny" ], "competences": [ "Fungal culturing and isolation", "Dual-culture and VOC assays", "Microscopy", "Genome assembly and annotation", "Comparative phylogenetics" ], "estimated_cost": "€8k-15k", "estimated_duration": "2-3 months" }, "expected_outputs": [ "Output 1: In vitro antagonism matrix (24 species × 2 pathogens × 4 traits, 4 replicates)", "Output 2: Draft genomes and BGC catalogues for 24 strains", "Output 3: Re-estimated Pagel's λ on 24-species subset", "Output 4: Cross-validated R² for phylogenetic vs random prediction of 12 untested species", "Output 5: Go/no-go decision document for Phase 3" ], "success_criteria": [ { "metric": "Pagel's λ on 24-species subset", "threshold": "λ ≥ 0.3, p < 0.05", "measurement": "phytools::phylosig on dual-culture inhibition data" }, { "metric": "Phylogenetic prediction R² vs random on 12 untested species", "threshold": "R²_phylo ≥ 0.4 and R²_random ≤ 0.1, p < 0.05", "measurement": "phylolm cross-validation, Wilcoxon signed-rank" }, { "metric": "Correlation between in vitro inhibition and BGC count", "threshold": "Spearman ρ ≥ 0.4, p < 0.05", "measurement": "antiSMASH BGC count vs mean inhibition %" }, { "metric": "Assay reproducibility", "threshold": "CV < 15% across 4 replicates", "measurement": "Coefficient of variation on inhibition %" } ], "go_nogo_decision": { "go_if": "λ ≥ 0.3 AND R²_phylo ≥ 0.4 AND R²_phylo - R²_random ≥ 0.3 AND BGC-inhibition ρ ≥ 0.4", "nogo_if": "λ < 0.1 OR R²_phylo ≤ 0.1 OR no correlation between in vitro and BGC content", "pivot_if": "λ 0.1-0.3 or R²_phylo 0.1-0.4: pivot to BGC-content-based prediction (drop phylogenetic weighting) and re-design Phase 3 SynComs using BGC profiles instead of phylogeny" }, "risks": [ { "risk": "Strains unavailable from culture collections or fail to grow", "probability": "medium", "mitigation": "Pre-order from 3 collections; include field isolation backup; use cryopreserved working stocks" }, { "risk": "In vitro antagonism does not correlate with in planta efficacy (key assumption)", "probability": "medium", "mitigation": "Include a pilot in planta check on 4 best/worst in vitro performers before full Phase 3; if no correlation, pivot to in planta-based screening" }, { "risk": "BGCs silent in vitro (epigenetic silencing)", "probability": "medium", "mitigation": "Add chemical elicitors (5-azacytidine, suberoylanilide hydroxamic acid) in a parallel assay; use RT-qPCR on key BGC genes" }, { "risk": "Contamination or cross-contamination in dual-culture", "probability": "low", "mitigation": "Use sealed I-plates, UV-sterilized hood, and ITS barcoding of all strains before and after assays" } ] }, { "phase_number": 3, "phase_name": "Full Experimental Protocol", "objective": "Rigorously validate that phylogenetically guided SynComs suppress plant disease more effectively than random SynComs of equivalent richness, across multiple pathogens and an agricultural intensification gradient, producing a publishable, generalizable result.", "methodology": "1) Expand to ≥100 endophyte species (from Phase 1 + new isolates) with full in vitro phenotyping (dual-culture, VOC, mycoparasitism, antibiosis) and draft genomes (antiSMASH BGC catalogues). 2) Build final time-calibrated phylogeny and estimate Pagel's λ per trait × pathogen. 3) Design SynComs: (a) phylogenetically guided (top predicted antagonists from distinct clades, maximizing phylogenetic diversity), (b) random (equivalent richness, random species), (c) monocultures of top predicted species, (d) untreated control. Richness levels: 3, 6, 12 species. 4) Greenhouse trials: wheat (F. graminearum, R. solani), tomato (B. cinerea), rice (M. oryzae) — susceptible genotypes, single pathogen strain per trial, 6 blocks, 10 plants per treatment. Measure disease index (0-100) at 14 dpi, lesion area (mm²), pathogen load by qPCR (pathogen DNA per ng plant DNA), and plant biomass. 5) Field trials (2 sites × 2 years) along an agricultural intensification gradient (low/medium/high, score 1-3): survey endophyte diversity, Faith's PD, and in vitro antagonism; test SynCom efficacy in plots. 6) Statistical models: linear mixed-effects (lme4/nlme) with SynCom type, richness, and intensification as fixed effects, block/site/year as random; ANOVA + Tukey; Spearman correlation for intensification vs PD/antagonism. 7) Stability: SynCom persistence in rhizosphere/endosphere by ITS amplicon sequencing (Illumina MiSeq) at 0, 14, 30, 60 dpi. 8) Pre-register on OSF; publish data and code on Zenodo/GitHub.", "required_resources": { "equipment": [ "Greenhouse with climate control (22-25°C, 60-70% RH, 12h light)", "Field trial sites (2 locations, 3 intensification levels)", "qPCR instrument", "Illumina MiSeq/NovaSeq (amplicon + genome)", "Plate reader, microscope, laminar hoods", "Plant growth chambers" ], "software": [ "R (lme4, nlme, phylolm, phytools, vegan, phyloseq)", "QIIME 2 / DADA2 for amplicon", "antiSMASH 7", "ImageJ", "SAS or R for mixed models" ], "datasets": [ "≥100 endophyte strains with genomes and phenotypes", "Wheat, tomato, rice susceptible genotypes", "F. graminearum PH-1, B. cinerea B05.10, R. solani AG1-IA, M. oryzae Guy11", "Field survey metadata (intensification index, soil, climate)" ], "competences": [ "Greenhouse and field experimentation", "Plant pathology (disease scoring, qPCR)", "Microbiome amplicon analysis", "Mixed-effects modeling", "SynCom design and assembly" ], "estimated_cost": "€60k-200k", "estimated_duration": "12-18 months" }, "expected_outputs": [ "Output 1: Full trait + genome + phylogeny dataset for ≥100 endophytes (public release)", "Output 2: Validated predictive model (phylogeny + BGC + mechanism) with cross-validated R²", "Output 3: Greenhouse and field SynCom efficacy data (predicted vs random vs monoculture vs control)", "Output 4: SynCom persistence and stability data (amplicon time series)", "Output 5: Intensification vs endophyte PD/antagonism correlation dataset", "Output 6: Peer-reviewed publication + pre-registered protocol + open data/code" ], "success_criteria": [ { "metric": "SynCom disease reduction (predicted vs random)", "threshold": "Predicted SynCom ≥ 30 percentage points more reduction than random, 95% CI excluding 0, p < 0.01", "measurement": "Linear mixed-effects model (lme4) with block as random effect, ANOVA + Tukey" }, { "metric": "Phylogenetic signal across full dataset", "threshold": "Pagel's λ ≥ 0.3 for ≥2 traits × ≥2 pathogens, p < 0.01", "measurement": "phytools::phylosig on ≥100 species" }, { "metric": "Predictive model R²", "threshold": "Cross-validated R² ≥ 0.4 for phylogenetic + BGC model vs R² ≤ 0.1 random", "measurement": "phylolm + cross-validation on held-out species" }, { "metric": "Intensification vs endophyte PD/antagonism", "threshold": "Spearman ρ ≤ -0.5, p < 0.01", "measurement": "Spearman correlation with multiple-comparison correction (Benjamini-Hochberg)" }, { "metric": "SynCom persistence", "threshold": "≥ 50% of introduced species detected at 60 dpi", "measurement": "ITS amplicon sequencing (QIIME 2/DADA2)" } ], "go_nogo_decision": { "go_if": "Predicted SynCom ≥ 30 pp better than random in ≥2 pathosystems AND model R² ≥ 0.4 AND λ ≥ 0.3 AND intensification ρ ≤ -0.5", "nogo_if": "Predicted SynCom not different from random (Δ ≤ 10 pp) OR model R² ≤ 0.1 OR λ < 0.1", "pivot_if": "Predicted SynCom 10-30 pp better than random: pivot to optimizing SynCom assembly rules (BGC complementarity, niche overlap) rather than pure phylogenetic selection; publish as proof-of-concept with revised model" }, "risks": [ { "risk": "Greenhouse results do not translate to field (environmental variability)", "probability": "high", "mitigation": "Include 2 sites × 2 years; use climate-controlled greenhouse as bridge; model environment × SynCom interaction" }, { "risk": "SynCom species incompatible or unstable in co-culture", "probability": "medium", "mitigation": "Pre-screen pairwise compatibility in vitro; use culturable, fast-growing strains; include persistence monitoring" }, { "risk": "Pathogen strain variability in field confounds results", "probability": "medium", "mitigation": "Use single characterized strain per trial; genotype pathogen populations by qPCR/amplicon; include strain as covariate" }, { "risk": "Regulatory or biosafety constraints for field release of fungi", "probability": "medium", "mitigation": "Use native, non-GMO endophytes; obtain permits early; conduct contained field trials first" }, { "risk": "Host genotype × SynCom interaction masks effect", "probability": "medium", "mitigation": "Use 2 susceptible genotypes per crop; include genotype as fixed effect; pre-test host susceptibility" }, { "risk": "Publication bias or reviewer resistance to phylogenetic prediction", "probability": "low", "mitigation": "Pre-register on OSF; publish negative results; share data/code on Zenodo/GitHub" } ] } ], "phase_1_quick_start": { "can_start_today": true, "first_action": "Download the 8 evidence-base papers and extract all reported antagonism trait values (species, pathogen, assay type, inhibition %) into a standardized CSV; simultaneously query NCBI GenBank for ITS/LSU/tef1/rpb2 sequences of the listed species and their close relatives.", "tools_needed": [ "R 4.4+ with ape, phytools, caper, phylolm", "IQ-TREE 2", "MAFFT v7", "antiSMASH 7", "Zotero or Rayyan for literature mining", "NCBI GenBank / Entrez Direct" ], "open_data_sources": [ "NCBI GenBank (https://www.ncbi.nlm.nih.gov/genbank/)", "MIBiG 3.0 (https://mibig.secondarymetabolites.org/)", "antiSMASH-DB (https://antismash-db.secondarymetabolites.org/)", "TreeBASE (https://treebase.org/)", "Fungal Trait Database (https://www.fungaltraits.org/)", "GlobalFungi (https://globalfungi.com/)", "OSF Preprints (https://osf.io/)" ] } }, "panel": { "reviews": [ { "reviewer_persona": "methodologist", "overall_score": 7.5, "verdict": "accept", "strengths": [ "Protocole structuré en trois phases avec des critères GO/NO-GO et des seuils quantitatifs explicites (λ ≥ 0,3, R² ≥ 0,4, etc.), ce qui facilite la reproductibilité et limite les décisions ad hoc.", "Pré-enregistrement sur OSF et publication des données et du code sur Zenodo/GitHub : démarche exemplaire pour la transparence et la reproductibilité.", "Prise en compte explicite du transfert horizontal de gènes (HGT) comme facteur confondant potentiel de la signalétique phylogénétique, avec une quantification prévue via antiSMASH/BiG-SLiCE.", "Plan de validation multi-échelle (in silico, in vitro, in planta, champ) et multi-pathogènes (F. graminearum, B. cinerea, R. solani, M. oryzae), ce qui augmente la validité externe.", "Utilisation de méthodes statistiques adaptées (Pagel's λ, Blomberg's K, phylolm, LRT, modèles mixtes) et d'une validation croisée pour comparer prédiction phylogénétique vs aléatoire." ], "weaknesses": [ "La puissance statistique n'est jamais estimée a priori pour les tests clés (Pagel's λ, comparaison de R², différence de réduction de maladie entre SynComs). Aucun calcul d'effectif requis ni de simulation de puissance n'est fourni, alors que les seuils (λ ≥ 0,3, Δ = 30 points de pourcentage) sont ambitieux et dépendent fortement de la taille d'échantillon et de la variabilité.", "Les contrôles expérimentaux sont incomplets : en Phase 3, il manque un contrôle « SynCom aléatoire avec mêmes espèces mais sans prédiction phylogénétique » (déjà partiellement présent) et surtout un contrôle « SynCom construit sur la base de la seule diversité phylogénétique sans prédiction de trait », ce qui empêche de distinguer l'effet de la phylogénie de celui de la diversité fonctionnelle.", "Le biais de publication est identifié comme risque mais aucune correction quantitative n'est prévue : la base de traits de Phase 1 sera inévitablement biaisée vers les genres bien étudiés (Trichoderma, Clonostachys), ce qui peut artificiellement gonfler le signal phylogénétique et fausser la validation croisée. Une analyse de sensibilité ou une pondération par l'effort d'étude est absente.", "La mesure de la « mycoparasitism incidence » par pourcentage de sites de contact avec enroulement/lyse est sujette à un biais de mesure subjectif (observateur-dépendant) sans évaluation de la répétabilité inter-observateurs. De même, le score d'antagonisme in vitro par inhibition radiale peut être confondu par des différences de vitesse de croissance intrinsèque.", "La Phase 3 prévoit des essais au champ sur 2 sites × 2 ans, mais la puissance pour détecter une différence de 30 points de pourcentage avec 6 blocs et 10 plantes par traitement n'est pas justifiée. De plus, l'analyse des communautés par amplicon ITS ne permet pas de suivre les souches introduites individuellement, ce qui rend le critère de persistance (≥ 50 % d'espèces détectées) peu fiable sans marquage spécifique.", "L'hypothèse de départ postule que le signal phylogénétique des traits antagonistes prédit l'efficacité in planta, mais le protocole ne contrôle pas la corrélation entre traits in vitro et efficacité in planta, qui est une condition nécessaire. Aucune analyse de médiation ou de corrélation n'est planifiée pour tester cette hypothèse clé avant les essais SynCom." ], "critical_questions": [ "Quelle est la puissance statistique estimée a priori pour détecter une différence de 30 points de pourcentage dans la réduction de maladie entre SynComs phylogénétiques et aléatoires, compte tenu de la variabilité inter-blocs, inter-sites et inter-années ? Un calcul d'effectif basé sur des données pilotes ou des simulations est-il prévu ?", "Comment le protocole contrôle-t-il le biais de publication dans la base de traits de Phase 1 ? Une analyse de sensibilité excluant les genres surreprésentés (Trichoderma, Clonostachys) ou une pondération par le nombre d'études publiées est-elle envisagée pour vérifier que le signal λ ≥ 0,3 n'est pas un artefact de suréchantillonnage taxonomique ?", "Quels contrôles supplémentaires sont prévus pour distinguer l'effet de la sélection phylogénétique de celui de la simple diversité phylogénétique ou fonctionnelle ? Par exemple, un traitement « SynCom aléatoire mais maximisant la diversité phylogénétique » est-il inclus pour isoler l'effet de la prédiction de trait ?", "Comment la répétabilité et la reproductibilité des mesures de mycoparasitisme (pourcentage de sites de contact avec enroulement/lyse) et de inhibition radiale sont-elles assurées ? Un test inter-observateurs et une standardisation de la taille de l'inoculum et de la phase de croissance sont-ils prévus pour éviter les biais de mesure ?", "La détection des espèces introduites dans les SynComs par séquençage ITS à 60 jours est-elle suffisamment résolutive pour distinguer les souches introduites des espèces indigènes ? Un marquage génétique ou des souches auxotrophes sont-ils envisagés pour un suivi fiable de la persistance ?" ], "recommendation": "Le protocole est globalement solide, bien structuré et intègre des bonnes pratiques (pré-enregistrement, critères GO/NO-GO, analyses phylogénétiques robustes). Cependant, il manque une analyse de puissance a priori, une gestion quantitative du biais de publication, des contrôles pour isoler l'effet de la prédiction phylogénétique de celui de la diversité, et une validation de la répétabilité des mesures subjectives. Je recommande une révision majeure pour intégrer ces éléments avant de considérer le protocole comme pleinement rigoureux. En l'état, je l'accepte avec des réserves substantielles.", "confidence": 0.85 }, { "reviewer_persona": "domain_expert", "overall_score": 6.5, "verdict": "weak_accept", "strengths": [ "L'hypothèse s'appuie sur un cadre théorique solide et bien établi : le phylogenetic niche conservatism et la conservation phylogénétique des traits microbiens (Aguirre de Cárcer 2019). L'idée d'étendre ce cadre à l'antagonisme fongique contre les phytopathogènes est conceptuellement cohérente et représente une avancée logique par rapport à la littérature existante.", "Le mécanisme proposé est articulé en une chaîne causale claire, allant des gènes (BGCs, effecteurs) aux phénotypes (antagonisme in vitro et in planta), puis à l'assemblage de communautés synthétiques. Cette progression est rationnelle et testable, ce qui est un point fort majeur.", "L'intégration de multiples niveaux d'analyse (génomique, phénotypique, écologique, agronomique) et la proposition d'un modèle prédictif combinant phylogénie, contenu en BGCs et classe de mécanisme est ambitieuse et potentiellement très utile pour le biocontrôle.", "La prise en compte explicite des incertitudes (taux de HGT, régulation épigénétique, stabilité des SynComs) montre une réflexion mature sur les limites du modèle." ], "weaknesses": [ "L'hypothèse suppose que le signal phylogénétique (Pagel's λ ≥ 0.3) est suffisamment fort et cohérent à travers l'ensemble de l'arbre fongique. Or, la littérature montre que les traits antagonistes peuvent être soumis à une forte sélection diversifiante ou à des transferts horizontaux de gènes (HGT), ce qui pourrait réduire considérablement le signal phylogénétique. Le seuil de λ ≥ 0.3 est arbitraire et n'est pas justifié par des données empiriques.", "La chaîne causale néglige l'importance des interactions environnementales et des facteurs biotiques (compétition, parasitisme, prédation) qui peuvent modifier l'expression des traits antagonistes. L'hypothèse que les conditions environnementales sont standardisées et n'affectent pas différentiellement la croissance des endophytes est irréaliste, en particulier in planta.", "Le lien entre antagonisme in vitro et efficacité in planta est loin d'être systématique. De nombreux exemples montrent que des souches très antagonistes in vitro échouent in planta, en raison de la colonisation, de la persistance ou de l'induction de résistance chez la plante. L'hypothèse ne traite pas de cette limite.", "La base bibliographique est relativement pauvre en études quantitatives sur le signal phylogénétique des traits antagonistes chez les champignons endophytes. La plupart des références citées sont des cas d'étude (Trichoderma, Xylaria) ou des revues conceptuelles, sans méta-analyse ou test formel de signal phylogénétique. Cela affaiblit la plausibilité empirique de l'hypothèse.", "L'hypothèse de HGT < 10% de la variance des traits est présentée comme une 'connaissance inconnue' mais elle est cruciale. Si le taux de HGT est plus élevé, la prédiction phylogénétique devient inefficace. Aucune méthode n'est proposée pour estimer ce taux ou pour l'intégrer dans le modèle." ], "critical_questions": [ "Quelle est la robustesse du signal phylogénétique (Pagel's λ) pour les traits antagonistes lorsque l'on contrôle pour la taille de l'échantillon, la couverture taxonomique et les erreurs de mesure ? Existe-t-il des clades fongiques où le signal est significativement plus faible, et pourquoi ?", "Comment l'hypothèse gère-t-elle les cas où des gènes de BGCs antagonistes sont acquis par HGT entre des lignées distantes ? Le modèle proposé peut-il distinguer la conservation verticale de l'acquisition horizontale, et comment cela affecte-t-il la prédiction ?", "Dans quelle mesure l'antagonisme in vitro est-il un prédicteur fiable de l'efficacité in planta ? Existe-t-il des données quantitatives sur la corrélation entre les deux, et comment l'hypothèse intègre-t-elle les facteurs propres à la plante (ex. induction de résistance, compétition avec le microbiome natif) ?", "Le seuil de λ ≥ 0.3 est-il justifié par des considérations théoriques ou empiriques ? Une valeur plus faible (ex. λ = 0.1) pourrait-elle encore permettre une prédiction utile, et comment le modèle s'ajuste-t-il à différents niveaux de signal phylogénétique ?", "Comment l'hypothèse traite-t-elle la régulation épigénétique des BGCs silencieux in vitro ? Si une grande partie des BGCs n'est pas exprimée dans les conditions de laboratoire, la mesure de l'antagonisme in vitro pourrait ne pas refléter le potentiel réel, ce qui biaiserait l'estimation du signal phylogénétique." ], "recommendation": "Je recommande une acceptation faible (weak_accept) car l'hypothèse est conceptuellement intéressante et bien articulée, avec un potentiel applicatif réel. Cependant, elle souffre de plusieurs faiblesses empiriques et théoriques : le seuil de signal phylogénétique est arbitraire, le rôle du HGT est sous-estimé, et le lien entre in vitro et in planta est simplifié. Pour renforcer la proposition, il serait nécessaire d'inclure une analyse de sensibilité du signal phylogénétique à travers différents clades et mécanismes, de quantifier le taux de HGT pour les BGCs antagonistes, et de valider la corrélation in vitro-in planta sur un large jeu de données. Une révision majeure est requise avant de considérer cette hypothèse comme pleinement soutenue.", "confidence": 0.8 }, { "reviewer_persona": "contrarian", "overall_score": 3.5, "verdict": "weak_reject", "strengths": [ "L'hypothèse est falsifiable et les auteurs spécifient des valeurs seuils (λ ≥ 0.3, R² ≥ 0.4, réduction de maladie ≥ 30%) qui permettent des tests statistiques clairs.", "L'approche intégrative combinant phylogénie, génomique des BGCs et essais in planta est ambitieuse et, si elle fonctionnait, aurait un impact appliqué réel en lutte biologique." ], "weaknesses": [ "FAIL REASON #1: Le lien entre antagonisme in vitro et efficacité in planta est notoirement faible et dépendant du contexte. De nombreux endophytes montrent une forte inhibition en dual-culture mais échouent à coloniser la plante ou à exprimer leurs BGCs in planta. L'hypothèse suppose une corrélation forte (non testée) entre les deux, ce qui invalide la chaîne causale des étapes 1 à 6. Sans validation préalable de cette corrélation sur un jeu de données indépendant, les prédictions 1, 2 et 3 peuvent être statistiquement significatives in vitro mais sans pertinence agricole.", "FAIL REASON #2: Le signal phylogénétique (Pagel's λ) peut être artificiellement gonflé par des confounders non contrôlés : (i) biais de mesure lié à la facilité de culture in vitro, qui est elle-même phylogénétiquement conservée ; (ii) effet de la taille du génome ou du nombre de BGCs, qui varie fortement entre clades et corrèle avec l'antagonisme sans lien causal avec la phylogénie ; (iii) HGT de BGCs entiers, dont le taux chez les champignons endophytes est inconnu et pourrait dépasser 10% de la variance, créant un signal phylogénétique apparent mais non vertical. L'hypothèse exclut explicitement ce scénario sans preuve.", "FAIL REASON #3: La puissance statistique pour détecter un effet de taille réaliste est probablement insuffisante. Avec ≥100 espèces, l'intervalle de confiance de λ est large (souvent ±0.2), et l'hypothèse H0: λ=0 est presque toujours rejetée avec un tel échantillon même pour un signal trivial (λ=0.1). De plus, la comparaison de R² entre prédiction phylogénétique et aléatoire (prédiction 2) est biaisée : la prédiction aléatoire devrait être évaluée par permutation, pas par un simple R²≤0.1. Enfin, pour l'essai in planta (prédiction 3), la variabilité des serres et l'effet lot rendent un différentiel de 30 points de pourcentage difficile à détecter avec un nombre réaliste de répétitions (n<10), menant à des faux négatifs ou à des faux positifs par sur-ajustement." ], "critical_questions": [ "Quelle est la corrélation attendue entre l'antagonisme in vitro (dual-culture) et l'efficacité in planta, et comment l'hypothèse tient-elle si cette corrélation est inférieure à 0.3, ce qui est fréquent dans la littérature ?", "Comment les auteurs contrôlent-ils le fait que la phylogénie des endophytes est corrélée à celle des plantes hôtes, et que l'antagonisme mesuré in vitro pourrait refléter une adaptation à l'hôte plutôt qu'une conservation phylogénétique des traits antagonistes ?", "Si l'HGT de BGCs est aussi fréquent que chez les bactéries (jusqu'à 20% pour certains clusters), quelle est la probabilité que le signal phylogénétique observé soit un artefact de transferts non verticaux, et comment l'hypothèse peut-elle être sauvée ?", "Les auteurs affirment que la force de l'antagonisme dépend de la distance phylogénétique entre endophyte et pathogène. Or, la prédiction 1 ne teste que l'antagonisme contre Fusarium graminearum. Comment généraliser à d'autres pathogènes sans introduire de biais de sélection ?" ], "recommendation": "Avant de tester le signal phylogénétique, les auteurs doivent démontrer sur un jeu de données indépendant que l'antagonisme in vitro prédit l'efficacité in planta avec un R² ≥ 0.5, et quantifier le taux d'HGT des BGCs antagonistes. Ensuite, ils doivent utiliser des modèles phylogénétiques explicites (PGLS) contrôlant pour la taille du génome, le nombre de BGCs et l'hôte, et valider la prédiction par validation croisée sur des clades entiers (pas seulement des espèces). Enfin, l'essai SynCom doit inclure un contrôle négatif de diversité équivalente mais sans endophytes antagonistes, et une analyse de puissance a priori pour justifier la taille d'échantillon.", "confidence": 0.85 }, { "reviewer_persona": "industrialist", "overall_score": 6.5, "verdict": "weak_accept", "strengths": [ "Le marché des biopesticides et biofertilisants microbiens croît à un CAGR de 12-15% et pèsera 10-15 milliards d'euros d'ici 2030. Des acteurs comme Bayer (acquisition de Ginkgo Bioworks), Syngenta (Biologicals), Corteva (avec l'accord sur les endophytes) et des pure-players comme Pivot Bio, BioConsortia ou Lallemand Plant Care paieraient pour un pipeline prédictif qui réduit le coût de découverte de nouveaux consortia synthétiques. Le coût actuel de screening d'un consortium efficace est de 500k-2M€ ; une méthode qui prédit l'efficacité avec R²≥0.4 et λ≥0.3 pourrait réduire ce coût de 40-60%.", "L'avantage compétitif est double : (1) un moteur de prédiction propriétaire basé sur la phylogénie et les biosynthetic gene clusters (BGC) qui crée un avantage de données cumulatif (plus on teste d'espèces, plus le modèle s'améliore) ; (2) une propriété intellectuelle sur les combinaisons SynCom prédites, pas seulement sur les souches individuelles. Les concurrents qui criblent au hasard ne peuvent pas rattraper cet avantage sans investir dans une base de données phylogénétique équivalente." ], "weaknesses": [ "La barrière à l'entrée est faible pour les concurrents qui disposent déjà de collections de souches et de pipelines de screening à haut débit (BASF, Bayer, Syngenta). La phylogénie n'est pas brevetable en soi ; seule l'application à des clades spécifiques et les SynComs résultants peuvent l'être. Le protocole proposé (18k-120k€) est un budget de recherche académique, pas de développement industriel : il ne couvre ni l'échelle, ni la réglementation, ni la formulation.", "Le risque commercial majeur est que le signal phylogénétique soit trop faible ou trop clade-spécifique pour être généralisable. Si λ<0.3 dans la plupart des clades, la prédiction ne bat pas le hasard, et l'ensemble de la proposition s'effondre. De plus, la réglementation des biopesticides (EPA, EFSA) exige des données d'efficacité au champ sur plusieurs années, ce qui allonge la timeline de 3-5 ans après la validation expérimentale. Enfin, les endophytes fongiques peuvent avoir des effets non ciblés sur le microbiome natif, ce qui complique l'homologation." ], "critical_questions": [ "Quel est le modèle économique exact : vendez-vous une plateforme de prédiction (SaaS) à des industriels, des SynComs prêts à l'emploi, ou des licences de brevets ? Le TAM pour une plateforme est de 50-100M€, contre 1-2Md€ pour des produits formulés, mais le coût de développement réglementaire est 10x plus élevé.", "Comment protégez-vous la propriété intellectuelle sur les prédictions phylogénétiques ? Les arbres phylogénétiques et les génomes fongiques sont souvent dans le domaine public. Sans brevets sur des séquences ou des combinaisons spécifiques, un concurrent peut reproduire votre approche en 6 mois avec les mêmes données ouvertes.", "Quel partenaire industriel accepterait de co-développer avec un TRL 3-4 et un budget de 120k€ ? Les grands comptes exigent généralement un TRL 6-7 et des données de champ avant de s'engager. Avez-vous identifié un early adopter prêt à financer la phase 3 (200k€) ?" ], "recommendation": "Concentrez-vous d'abord sur la phase 1 in silico (4-8 semaines, 2k€) pour vérifier que λ≥0.3 et R²≥0.4 sur au moins 2 traits et 2 pathogènes. Si c'est validé, déposez un brevet provisoire sur la méthode de sélection phylogénétique appliquée à un clade fongique spécifique (ex. Trichoderma, Beauveria) et cherchez un partenariat avec un acteur de taille moyenne (Lallemand, Koppert, Biobest) plutôt qu'avec Bayer ou Syngenta, qui voudront tout internaliser. La phase 3 (200k€) doit être cofinancée par un industriel ou un programme européen (Horizon Europe, EIC Accelerator) pour valider l'efficacité in planta sur blé contre Fusarium. Sans partenaire industriel à ce stade, le projet reste académique et le ROI est nul.", "confidence": 0.65 }, { "reviewer_persona": "funding_strategist", "overall_score": 6.5, "verdict": "weak_accept", "strengths": [ "Hypothèse originale combinant phylogénie fongique, génomique des BGC et écologie synthétique, avec un fort potentiel de généralisation au-delà des endophytes.", "Protocole en trois phases avec critères GO/NO-GO quantitatifs et budget modulaire (18–120 k€), ce qui facilite l'ancrage dans des appels à faible TRL et à haut risque.", "Positionnement à l'interface entre lutte biologique et biologie synthétique communautaire, un créneau peu saturé et attractif pour les évaluateurs européens." ], "weaknesses": [ "Le TRL actuel est très bas (TRL 2–3) : la phase in silico repose sur des données de traits antagonistes souvent hétérogènes et peu standardisées, ce qui fragilise la reproductibilité.", "Le consortium n'est pas défini et l'hypothèse exige une double compétence rare (phylogenomics fongique + pathologie végétale + SynComs), difficile à réunir dans un seul laboratoire.", "La valeur seuil λ ≥ 0,3 est arbitraire et pourrait être rejetée par des évaluateurs exigeants sans justification théorique ou simulation de puissance préalable." ], "critical_questions": [ "Disposez-vous déjà d'un jeu de données phénotypiques d'antagonisme suffisamment standardisé (≥100 souches, ≥2 pathogènes) pour justifier la phase in silico sans collecte préalable ?", "Quel est le plan de gestion des conflits d'intérêts et de propriété intellectuelle si un partenaire industriel (biocontrôle) rejoint le consortium en phase 3 ?", "Comment garantissez-vous que le signal phylogénétique observé in vitro se transpose in planta, sachant que les traits antagonistes sont fortement modulés par l'hôte et le sol ?" ], "recommendation": "Ciblez d'abord un appel ANR à faible TRL (JCJC ou PRC) pour financer les phases 1–2 et produire une preuve de concept publiable, puis capitalisez sur ces résultats pour viser un ERC Starting Grant ou un appel Horizon Europe (HORIZON-CL6) en phase 3. Évitez de soumettre directement en ERC sans données préliminaires : le risque de rejet pour manque de faisabilité est trop élevé.", "confidence": 0.72, "funding_programs": [ { "program": "ANR JCJC 2026 (Jeunes Chercheuses et Jeunes Chercheurs)", "agency": "Agence Nationale de la Recherche (ANR)", "fit_score": 0.85, "typical_budget": "200–400 k€ sur 36–48 mois", "success_rate": "≈ 15–20 % selon les éditions", "next_deadline": "Octobre 2025 (pré-projet) / Mars 2026 (projet complet)", "rationale": "L'ANR JCJC est idéale pour un porteur unique souhaitant financer les phases 1 et 2 (in silico + validation expérimentale minimale) avec un budget modeste et un TRL 2–3. Le format encourage la prise de risque et l'originalité, ce qui correspond exactement à l'hypothèse du signal phylogénétique comme prédicteur de l'efficacité des SynComs." }, { "program": "ERC Starting Grant 2026", "agency": "European Research Council (ERC)", "fit_score": 0.7, "typical_budget": "1,5–2 M€ sur 5 ans", "success_rate": "≈ 10–12 %", "next_deadline": "Octobre 2025 (lettre d'intention) / Janvier 2026 (proposition complète)", "rationale": "L'ERC Starting Grant finance des projets frontière à haut risque et à long terme, ce qui correspond à la vision générale de l'hypothèse (phylogénie comme cadre prédictif universel). Cependant, le TRL actuel est trop bas et le consortium non constitué : il faut d'abord des données préliminaires solides (phase 1–2) avant de soumettre, sinon le projet sera perçu comme spéculatif." }, { "program": "HORIZON-CL6-2026-02-BIODIV-01 : Biocontrol and integrated pest management for sustainable agriculture", "agency": "Commission Européenne (Horizon Europe, Cluster 6)", "fit_score": 0.75, "typical_budget": "3–6 M€ sur 36–48 mois (consortium de 6–10 partenaires)", "success_rate": "≈ 12–15 %", "next_deadline": "Février 2026 (appel à propositions)", "rationale": "Cet appel cible explicitement des solutions de biocontrôle innovantes et durables pour l'agriculture, avec une attente de démonstration in planta et de consortium multi-acteurs. La phase 3 du protocole (SynComs phylogénétiquement guidés, gradient d'intensification agricole) correspond parfaitement aux attendus, mais nécessite un consortium élargi incluant des partenaires industriels et des instituts agricoles." } ] } ], "meta_review": { "consensus_score": 6.06, "verdict": "publish_brief", "key_consensus": [ "Le protocole est globalement bien structuré, avec des critères GO/NO-GO quantitatifs, un pré-enregistrement et une démarche de transparence (données et code ouverts) qui sont salués par plusieurs reviewers.", "L'hypothèse est conceptuellement originale et articule une chaîne causale claire entre gènes (BGCs), phénotypes antagonistes et efficacité in planta, ce qui représente une avancée potentielle pour le biocontrôle.", "Le lien entre antagonisme in vitro et efficacité in planta est reconnu comme une condition nécessaire mais non testée, et son absence de validation préalable fragilise l'ensemble de la prédiction.", "Le seuil de signal phylogénétique (Pagel's λ ≥ 0,3) est jugé arbitraire et non justifié par des données empiriques ou une analyse de puissance, ce qui constitue une faiblesse partagée.", "Le transfert horizontal de gènes (HGT) est identifié comme un facteur confondant majeur qui pourrait invalider la prédiction phylogénétique s'il n'est pas quantifié." ], "key_disagreements": [ "Le methodologist et le domain_expert considèrent que le protocole est révisable et recommandent une révision majeure, tandis que le contrarian estime que les failles sont trop fondamentales (FAIL REASON #1, #2, #3) pour être corrigées sans une validation préalable sur données indépendantes.", "L'industrialist et le funding_strategist voient un potentiel applicatif et des opportunités de financement, mais soulignent que le TRL est trop bas et que le consortium n'est pas défini, ce qui contraste avec l'optimisme des reviewers académiques sur la faisabilité.", "Le domain_expert insiste sur la nécessité de valider la corrélation in vitro-in planta sur un large jeu de données avant de tester le signal phylogénétique, alors que le methodologist propose des contrôles supplémentaires mais ne remet pas en cause la séquence des phases.", "Le contrarian affirme que la puissance statistique est insuffisante pour détecter un effet réaliste et que la prédiction aléatoire est mal évaluée, tandis que le methodologist reconnaît l'absence d'analyse de puissance mais ne la juge pas rédhibitoire." ], "critical_path": "La validation préalable de la corrélation entre antagonisme in vitro et efficacité in planta sur un jeu de données indépendant, avec un R² ≥ 0,5, est le facteur le plus déterminant : sans cette preuve, la chaîne causale s'effondre et les prédictions perdent leur pertinence agricole. En parallèle, la quantification du taux de HGT des BGCs antagonistes et une analyse de puissance a priori sont indispensables pour établir la robustesse du signal phylogénétique.", "final_recommendation": "Le panel reconnaît l'originalité et la rigueur du protocole, mais les faiblesses identifiées sont trop fondamentales pour être résolues dans le cadre d'une révision. Le lien non validé entre antagonisme in vitro et efficacité in planta, le seuil arbitraire de λ ≥ 0,3, l'absence d'analyse de puissance et la sous-estimation du HGT compromettent la validité de l'hypothèse. Le contrarian, avec une confiance élevée, souligne que sans validation préalable sur données indépendantes, les prédictions pourraient être statistiquement significatives in vitro mais sans pertinence agricole. En conséquence, le panel recommande de rejeter l'hypothèse en l'état et d'inviter les auteurs à produire d'abord des preuves empiriques de la corrélation in vitro-in planta et une quantification du HGT avant de soumettre à nouveau.", "brief_quality_gate": false, "revision_guidance": [], "llm_verdict": "reject", "llm_consensus_score": 6.2, "verdict_override_reason": "Python threshold override: consensus 6.06 at iter 2 → publish_brief (LLM said reject)" } }, "vulgarization_fr": { "title_fr": "Champignons protecteurs : la parentèle prédit-elle l'efficacité ?", "hypothesis_in_brief": "Les champignons microscopiques qui vivent dans les plantes produisent souvent des substances capables de bloquer les champignons pathogènes. Cette hypothèse propose que ces capacités se transmettent par héritage évolutif : plus deux champignons sont proches parents, plus ils partagent les mêmes armes chimiques. Si c'est vrai, il suffirait d'analyser l'arbre généalogique des champignons pour prédire lesquels seront de bons gardes du corps, sans avoir à tous les tester au laboratoire.", "why_it_matters": "Aujourd'hui, pour trouver un champignon capable de protéger une culture, il faut tester des centaines de souches une par une, ce qui coûte cher et prend des années. Une méthode de prédiction fiable permettrait de cibler directement les espèces les plus prometteuses, réduisant fortement le coût de découverte de nouveaux traitements biologiques contre les maladies des plantes. Cela intéresserait les agriculteurs qui cherchent à réduire les pesticides chimiques, ainsi que les entreprises qui développent des produits de biocontrôle. À plus long terme, mieux comprendre comment l'intensification agricole appauvrit ces communautés protectrices pourrait aider à préserver ce service naturel.", "imagine_that": "Imaginez que vous cherchez un bon serrurier dans une ville inconnue. Plutôt que de tous les tester un par un, vous remarquez que les serruriers compétents ont souvent été formés par les mêmes maîtres, qui leur ont transmis les mêmes techniques. En reconstituant l'arbre des transmissions entre maîtres et élèves, vous pouvez deviner qui est compétent avant même de l'avoir vu travailler. Ici, l'idée est la même : les champignons héritent de leurs ancêtres des « recettes » chimiques pour bloquer les pathogènes, et en retraçant leur arbre généalogique, on peut prédire qui possède ces recettes sans avoir à les tester tous.", "concretely": { "intro": "L'approche se déroule en trois étapes, de la pure analyse de données existantes jusqu'à des essais en conditions agricoles.", "phase1": "Il s'agit d'assembler une grande base de données à partir de la littérature scientifique : des centaines de mesures d'inhibition entre champignons endophytes et pathogènes. Un modèle statistique teste ensuite si ces capacités se regroupent par familles évolutives, et si connaître la position d'une espèce dans l'arbre permet de prédire son efficacité mieux que le hasard.", "phase2": "Vingt-quatre espèces sont sélectionnées, dont la moitié déjà testées et l'autre moitié des proches parentes jamais évaluées. Des expériences de confrontation en boîte de Petri vérifient si les prédictions issues de la phase 1 se confirment sur ces nouvelles espèces.", "phase3": "L'échelle monte à plus de cent espèces, avec des tests complets et des essais sur plantes entières. Des communautés synthétiques construites selon les prédictions phylogénétiques sont comparées à des communautés aléatoires de même richesse, sur plusieurs pathogènes et le long d'un gradient d'intensification agricole." }, "reviewers_say": "Le panel reconnaît une hypothèse originale et un protocole bien structuré, avec des critères d'arrêt clairs et une démarche transparente de pré-enregistrement et de partage des données. La chaîne causale, qui va des gènes aux phénotypes puis à l'efficacité au champ, est jugée logique et testable. Cependant, plusieurs faiblesses sont pointées : le lien entre l'inhibition observée en boîte de Petri et la protection réelle sur la plante est notoirement faible et dépend du contexte, ce qui fragilise toute la prédiction. Le seuil de signal phylogénétique (λ ≥ 0,3) est jugé arbitraire, aucune analyse de puissance n'est fournie, et les transferts horizontaux de gènes pourraient gonfler artificiellement le signal. Le verdict global est de publier ce brief comme piste à explorer, mais pour y croire vraiment, il faudrait d'abord démontrer sur des données indépendantes que l'antagonisme in vitro prédit l'efficacité in planta, et estimer la puissance statistique des tests clés." }, "vulgarization_en": { "title": "Protective fungi: does kinship predict efficacy?", "hypothesis_in_brief": "Microscopic fungi living within plants often produce substances capable of blocking pathogenic fungi. This hypothesis proposes that such capacities are transmitted through evolutionary inheritance: the more closely related two fungi are, the more they share the same chemical weapons. If this holds true, it would suffice to analyse the genealogical tree of fungi to predict which will serve as effective bodyguards, without the need to test them all in the laboratory.", "why_it_matters": "Today, finding a fungus capable of protecting a crop requires testing hundreds of strains one by one, which is costly and takes years. A reliable prediction method would make it possible to target the most promising species directly, sharply reducing the cost of finding new biological treatments against plant diseases. This would interest farmers seeking to reduce chemical pesticides, as well as companies developing biocontrol products. In the longer term, a better understanding of how agricultural intensification depletes these protective communities could help preserve this natural service.", "imagine_that": "Imagine that you are looking for a good locksmith in an unfamiliar city. Rather than testing every one of them individually, you notice that competent locksmiths have often been trained by the same masters, who passed on the same techniques. By reconstructing the tree of transmissions between masters and apprentices, you can deduce who is competent before you have even seen them work. Here, the idea is the same: fungi inherit from their ancestors chemical \"recipes\" for blocking pathogens, and by tracing their family tree, you can predict which ones possess these recipes without having to test them all.", "reviewers_say": "The panel recognises an original hypothesis and a well-structured protocol, with clear stopping criteria and a transparent approach to pre-registration and data sharing. The causal chain, which runs from genes to phenotypes and then to field efficacy, is judged to be logical and testable. However, several weaknesses are identified: the link between the inhibition observed in Petri dishes and the actual protection on the plant is notoriously weak and context-dependent, which undermines the entire prediction. The phylogenetic signal threshold (λ ≥ 0.3) is judged to be arbitrary, no power analysis is provided, and horizontal gene transfers could artificially inflate the signal. The overall verdict is to publish this brief as a lead worth exploring, but to truly believe in it, it would first be necessary to demonstrate on independent data that in vitro antagonism predicts in planta efficacy, and to estimate the statistical power of the key tests.", "concretely": { "intro": "The approach proceeds in three stages, from pure analysis of existing data through to trials under agricultural conditions.", "phase1": "The task is to assemble a large database from the scientific literature: hundreds of inhibition measurements between endophytic fungi and pathogens. A statistical model then tests whether these capabilities cluster by evolutionary family, and whether knowing a species' position in the tree allows its efficacy to be predicted better than chance.", "phase2": "Twenty-four species are selected, half of which have already been tested and the other half being close relatives never evaluated. Confrontation experiments in Petri dishes verify whether the predictions from phase 1 are confirmed in these new species.", "phase3": "The scale extends to more than one hundred species, with full testing and whole-plant trials. Synthetic communities assembled according to phylogenetic predictions are compared with random communities of equal richness, across several pathogens and along a gradient of agricultural intensification." } }, "panel_en": { "reviews": [ { "reviewer_persona": "methodologist", "overall_score": 7.5, "verdict": "accept", "confidence": 0.85, "strengths": [ "A protocol structured in three phases with GO/NO-GO criteria and explicit quantitative thresholds (λ ≥ 0.3, R² ≥ 0.4, etc.), which facilitates reproducibility and limits ad hoc decisions.", "Pre-registration on OSF and publication of data and code on Zenodo/GitHub: an exemplary approach for transparency and reproducibility.", "Explicit consideration of horizontal gene transfer (HGT) as a potential confounder of phylogenetic signal, with quantification planned via antiSMASH/BiG-SLiCE.", "A multi-scale (in silico, in vitro, in planta, field) and multi-pathogen (F. graminearum, B. cinerea, R. solani, M. oryzae) validation plan, which increases external validity.", "Use of appropriate statistical methods (Pagel's λ, Blomberg's K, phylolm, LRT, mixed models) and of cross-validation to compare phylogenetic versus random prediction." ], "weaknesses": [ "Statistical power is never estimated a priori for the key tests (Pagel's λ, R² comparison, difference in disease reduction between SynComs). No sample-size calculation or power simulation is provided, even though the thresholds (λ ≥ 0.3, Δ = 30 percentage points) are ambitious and depend strongly on sample size and variability.", "The experimental controls are incomplete: in Phase 3, a \"random SynCom with the same species but without phylogenetic prediction\" control is missing (already partially present) and, above all, a \"SynCom constructed on the basis of phylogenetic diversity alone without trait prediction\" control is missing, which prevents the effect of phylogeny from being distinguished from that of functional diversity.", "Publication bias is identified as a risk, but no quantitative correction is planned: the Phase 1 trait database will inevitably be biased towards well-studied genera (Trichoderma, Clonostachys), which may artificially inflate the phylogenetic signal and distort cross-validation. A sensitivity analysis or weighting by study effort is absent.", "The measurement of \"mycoparasitism incidence\" as the percentage of contact sites with coiling/lysis is subject to subjective (observer-dependent) measurement bias, with no assessment of inter-observer repeatability. Likewise, the in vitro antagonism score based on radial inhibition may be confounded by differences in intrinsic growth rate.", "Phase 3 plans field trials across 2 sites × 2 years, but the power to detect a difference of 30 percentage points with 6 blocks and 10 plants per treatment is not justified. Moreover, community analysis by ITS amplicon does not allow the introduced strains to be tracked individually, which renders the persistence endpoint (≥ 50% of species detected) unreliable without specific tagging.", "The starting hypothesis posits that the phylogenetic signal of antagonistic traits predicts in planta efficacy, but the protocol does not control for the correlation between in vitro traits and in planta efficacy, which is a necessary condition. No mediation or correlation analysis is planned to test this key hypothesis before the SynCom trials." ], "critical_questions": [ "What is the statistical power estimated a priori to detect a 30 percentage-point difference in disease reduction between phylogenetic and random SynComs, given inter-block, inter-site and inter-year variability? Is a sample-size calculation based on pilot data or simulations planned?", "How does the protocol control for publication bias in the Phase 1 trait database? Is a sensitivity analysis excluding over-represented genera (Trichoderma, Clonostachys) or weighting by the number of published studies envisaged, to verify that the λ ≥ 0.3 signal is not an artefact of taxonomic oversampling?", "What additional controls are planned to distinguish the effect of phylogenetic selection from that of mere phylogenetic or functional diversity? For example, is a \"random SynCom maximising phylogenetic diversity\" treatment included to isolate the effect of trait prediction?", "How are the repeatability and reproducibility of mycoparasitism measurements (percentage of contact sites with coiling/lysis) and of radial inhibition ensured? Are an inter-observer test and standardisation of inoculum size and growth phase planned to avoid measurement bias?", "Is the detection of species introduced into the SynComs by ITS sequencing at 60 days sufficiently resolutive to distinguish introduced strains from indigenous species? Are genetic tagging or auxotrophic strains envisaged for reliable monitoring of persistence?" ], "recommendation": "The protocol is broadly sound, well structured and incorporates good practice (preregistration, GO/NO-GO criteria, robust phylogenetic analyses). However, it lacks an a priori power analysis, a quantitative treatment of publication bias, controls to isolate the effect of phylogenetic prediction from that of diversity, and validation of the repeatability of subjective measurements. A major revision is recommended to incorporate these elements before the protocol can be considered fully rigorous. As it stands, it is accepted with substantial reservations." }, { "reviewer_persona": "domain_expert", "overall_score": 6.5, "verdict": "weak_accept", "confidence": 0.8, "strengths": [ "The hypothesis rests on a robust and well-established theoretical framework: phylogenetic niche conservatism and the phylogenetic conservation of microbial traits (Aguirre de Cárcer 2019). The extension of this framework to fungal antagonism against phytopathogens is conceptually coherent and constitutes a logical advance over the existing literature.", "The proposed mechanism is articulated as a clear causal chain, proceeding from genes (BGCs, effectors) to phenotypes (antagonism in vitro and in planta), and thence to the assembly of synthetic communities. This progression is rational and testable, which is a major strength.", "The integration of multiple levels of analysis (genomic, phenotypic, ecological, agronomic) and the proposal of a predictive model combining phylogeny, BGC content and mechanism class is ambitious and potentially of considerable utility for biocontrol.", "The explicit consideration of uncertainties (HGT rate, epigenetic regulation, stability of SynComs) demonstrates a mature reflection on the limitations of the model." ], "weaknesses": [ "The hypothesis assumes that the phylogenetic signal (Pagel's λ ≥ 0.3) is sufficiently strong and consistent across the entire fungal tree. However, the literature shows that antagonistic traits may be subject to strong diversifying selection or to horizontal gene transfer (HGT), which could considerably reduce the phylogenetic signal. The threshold of λ ≥ 0.3 is arbitrary and is not justified by empirical data.", "The causal chain neglects the importance of environmental interactions and biotic factors (competition, parasitism, predation) that may modify the expression of antagonistic traits. The hypothesis that environmental conditions are standardised and do not differentially affect endophyte growth is unrealistic, particularly in planta.", "The link between in vitro antagonism and in planta efficacy is far from systematic. Numerous examples show that highly antagonistic strains in vitro fail in planta, owing to colonisation, persistence or the induction of resistance in the plant. The hypothesis does not address this limitation.", "The bibliographic base is relatively poor in quantitative studies on the phylogenetic signal of antagonistic traits in endophytic fungi. Most of the references cited are case studies (Trichoderma, Xylaria) or conceptual reviews, without meta-analysis or formal testing of phylogenetic signal. This weakens the empirical plausibility of the hypothesis.", "The hypothesis of HGT < 10% of trait variance is presented as an 'unknown known' but it is crucial. If the HGT rate is higher, the phylogenetic prediction becomes ineffective. No method is proposed to estimate this rate or to incorporate it into the model." ], "critical_questions": [ "What is the robustness of the phylogenetic signal (Pagel's λ) for antagonistic traits when sample size, taxonomic coverage and measurement error are controlled for? Are there fungal clades in which the signal is significantly weaker, and why?", "How does the hypothesis handle cases in which antagonistic BGC genes are acquired by HGT between distant lineages? Can the proposed model distinguish vertical inheritance from horizontal acquisition, and how does this affect prediction?", "To what extent is in vitro antagonism a reliable predictor of in planta efficacy? Are quantitative data available on the correlation between the two, and how does the hypothesis incorporate plant-specific factors (e.g. induction of resistance, competition with the native microbiome)?", "Is the threshold of λ ≥ 0.3 justified by theoretical or empirical considerations? Could a lower value (e.g. λ = 0.1) still permit useful prediction, and how does the model adjust to different levels of phylogenetic signal?", "How does the hypothesis address the epigenetic regulation of BGCs that are silent in vitro? If a large proportion of BGCs is not expressed under laboratory conditions, the measurement of in vitro antagonism may not reflect the true potential, which would bias the estimation of the phylogenetic signal." ], "recommendation": "A weak_accept is recommended, as the hypothesis is conceptually interesting and well articulated, with genuine applicative potential. However, it suffers from several empirical and theoretical weaknesses: the phylogenetic signal threshold is arbitrary, the role of HGT is underestimated, and the link between in vitro and in planta is oversimplified. To strengthen the proposal, it would be necessary to include a sensitivity analysis of the phylogenetic signal across different clades and mechanisms, to quantify the HGT rate for antagonistic BGCs, and to validate the in vitro–in planta correlation on a large dataset. Major revision is required before this hypothesis can be considered fully supported." }, { "reviewer_persona": "contrarian", "overall_score": 3.5, "verdict": "weak_reject", "confidence": 0.85, "strengths": [ "The hypothesis is falsifiable, and threshold values are specified by the authors (λ ≥ 0.3, R² ≥ 0.4, disease reduction ≥ 30%) that permit clear statistical testing.", "The integrative approach combining phylogeny, BGC genomics and in planta assays is ambitious and, if successful, would have a tangible applied impact in biological control." ], "weaknesses": [ "FAIL REASON #1: The link between in vitro antagonism and in planta efficacy is notoriously weak and context-dependent. Many endophytes exhibit strong inhibition in dual culture yet fail to colonise the plant or to express their BGCs in planta. The hypothesis assumes a strong (untested) correlation between the two, which invalidates the causal chain of steps 1 to 6. Without prior validation of this correlation on an independent dataset, predictions 1, 2 and 3 may be statistically significant in vitro yet of no agricultural relevance.", "FAIL REASON #2: The phylogenetic signal (Pagel's λ) may be artificially inflated by uncontrolled confounders: (i) measurement bias linked to ease of in vitro culture, which is itself phylogenetically conserved; (ii) the effect of genome size or of the number of BGCs, which varies strongly between clades and correlates with antagonism without any causal link to phylogeny; (iii) HGT of entire BGCs, the rate of which in endophytic fungi is unknown and could exceed 10% of the variance, creating an apparent but non-vertical phylogenetic signal. The hypothesis explicitly excludes this scenario without evidence.", "FAIL REASON #3: The statistical power to detect an effect of realistic size is probably insufficient. With ≥100 species, the confidence interval of λ is wide (often ±0.2), and the null hypothesis H0: λ=0 is almost always rejected with such a sample even for a trivial signal (λ=0.1). Moreover, the comparison of R² between phylogenetic and random prediction (prediction 2) is biased: the random prediction should be evaluated by permutation, not by a simple R²≤0.1. Finally, for the in planta trial (prediction 3), greenhouse variability and the batch effect make a differential of 30 percentage points difficult to detect with a realistic number of replicates (n<10), leading to false negatives or to false positives through overfitting." ], "critical_questions": [ "What is the expected correlation between in vitro antagonism (dual-culture) and in planta efficacy, and how does the hypothesis hold if this correlation is below 0.3, which is frequently reported in the literature?", "How do the authors control for the fact that endophyte phylogeny is correlated with that of host plants, and that the antagonism measured in vitro might reflect adaptation to the host rather than phylogenetic conservation of antagonistic traits?", "If HGT of BGCs is as frequent as in bacteria (up to 20% for certain clusters), what is the probability that the observed phylogenetic signal is an artefact of non-vertical transfers, and how can the hypothesis be rescued?", "The authors assert that the strength of antagonism depends on the phylogenetic distance between endophyte and pathogen. However, prediction 1 tests antagonism only against Fusarium graminearum. How can this be generalised to other pathogens without introducing selection bias?" ], "recommendation": "Before the phylogenetic signal is tested, it must be demonstrated on an independent dataset that in vitro antagonism predicts in planta efficacy with an R² ≥ 0.5, and the HGT rate of antagonistic BGCs must be quantified. The panel then requires the use of explicit phylogenetic models (PGLS) controlling for genome size, BGC count and host, with prediction validated by cross-validation across entire clades (not merely species). Finally, the SynCom trial must include a negative control of equivalent diversity but without antagonistic endophytes, together with an a priori power analysis to justify the sample size." }, { "reviewer_persona": "industrialist", "overall_score": 6.5, "verdict": "weak_accept", "confidence": 0.65, "strengths": [ "The microbial biopesticide and biofertiliser market is growing at a CAGR of 12–15% and will be worth €10–15 billion by 2030. Players such as Bayer (acquisition of Ginkgo Bioworks), Syngenta (Biologicals), Corteva (with the endophyte agreement) and pure-players such as Pivot Bio, BioConsortia and Lallemand Plant Care would pay for a predictive pipeline that reduces the cost of finding new synthetic consortia. The current cost of screening an effective consortium is €500k–2M; a method that predicts efficacy with R²≥0.4 and λ≥0.3 could reduce this cost by 40–60%.", "The competitive advantage is twofold: (1) a proprietary prediction engine based on phylogeny and biosynthetic gene clusters (BGC) that creates a cumulative data advantage (the more species are tested, the more the model improves); (2) intellectual property on the predicted SynCom combinations, not only on individual strains. Competitors that screen at random cannot close this gap without investing in an equivalent phylogenetic database." ], "weaknesses": [ "The barrier to entry is low for competitors that already possess strain collections and high-throughput screening pipelines (BASF, Bayer, Syngenta). Phylogeny is not patentable as such; only its application to specific clades and the resulting SynComs can be. The proposed protocol (€18k–120k) corresponds to an academic research budget, not industrial development: it covers neither scale-up, nor regulatory requirements, nor formulation.", "The principal commercial risk is that the phylogenetic signal may be too weak or too clade-specific to be generalisable. If λ<0.3 in most clades, prediction does not outperform chance, and the entire proposal collapses. Furthermore, biopesticide regulation (EPA, EFSA) requires multi-year field efficacy data, which extends the timeline by 3–5 years beyond experimental validation. Finally, fungal endophytes may exert off-target effects on the native microbiome, which complicates approval." ], "critical_questions": [ "What is the exact business model: is a prediction platform (SaaS) sold to industrial firms, ready-to-use SynComs, or patent licences? The TAM for a platform is €50–100M, against €1–2bn for formulated products, but the cost of regulatory development is 10× higher.", "How is intellectual property protected on phylogenetic predictions? Phylogenetic trees and fungal genomes are often in the public domain. Without patents on specific sequences or combinations, a competitor can reproduce the approach within 6 months using the same open data.", "Which industrial partner would agree to co-develop at TRL 3–4 with a budget of €120k? Large accounts generally require TRL 6–7 and field data before committing. Has an early adopter been identified who is prepared to fund Phase 3 (€200k)?" ], "recommendation": "Focus first on Phase 1 in silico (4–8 weeks, €2k) to verify that λ≥0.3 and R²≥0.4 across at least 2 traits and 2 pathogens. If this is validated, a provisional patent should be filed on the phylogenetic selection method applied to a specific fungal clade (e.g. Trichoderma, Beauveria), and a partnership should be sought with a mid-sized player (Lallemand, Koppert, Biobest) rather than with Bayer or Syngenta, which will want to internalise everything. Phase 3 (€200k) must be co-funded by an industrial partner or a European programme (Horizon Europe, EIC Accelerator) in order to validate in planta efficacy on wheat against Fusarium. Without an industrial partner at that stage, the project remains academic and the ROI is nil." }, { "reviewer_persona": "funding_strategist", "overall_score": 6.5, "verdict": "weak_accept", "confidence": 0.72, "funding_programs": [ { "program": "ANR JCJC 2026 (Jeunes Chercheuses et Jeunes Chercheurs)", "agency": "Agence Nationale de la Recherche (ANR)", "fit_score": 0.85, "typical_budget": "200–400 k€ sur 36–48 mois", "success_rate": "≈ 15–20 % selon les éditions", "next_deadline": "Octobre 2025 (pré-projet) / Mars 2026 (projet complet)", "rationale": "L'ANR JCJC est idéale pour un porteur unique souhaitant financer les phases 1 et 2 (in silico + validation expérimentale minimale) avec un budget modeste et un TRL 2–3. Le format encourage la prise de risque et l'originalité, ce qui correspond exactement à l'hypothèse du signal phylogénétique comme prédicteur de l'efficacité des SynComs." }, { "program": "ERC Starting Grant 2026", "agency": "European Research Council (ERC)", "fit_score": 0.7, "typical_budget": "1,5–2 M€ sur 5 ans", "success_rate": "≈ 10–12 %", "next_deadline": "Octobre 2025 (lettre d'intention) / Janvier 2026 (proposition complète)", "rationale": "L'ERC Starting Grant finance des projets frontière à haut risque et à long terme, ce qui correspond à la vision générale de l'hypothèse (phylogénie comme cadre prédictif universel). Cependant, le TRL actuel est trop bas et le consortium non constitué : il faut d'abord des données préliminaires solides (phase 1–2) avant de soumettre, sinon le projet sera perçu comme spéculatif." }, { "program": "HORIZON-CL6-2026-02-BIODIV-01 : Biocontrol and integrated pest management for sustainable agriculture", "agency": "Commission Européenne (Horizon Europe, Cluster 6)", "fit_score": 0.75, "typical_budget": "3–6 M€ sur 36–48 mois (consortium de 6–10 partenaires)", "success_rate": "≈ 12–15 %", "next_deadline": "Février 2026 (appel à propositions)", "rationale": "Cet appel cible explicitement des solutions de biocontrôle innovantes et durables pour l'agriculture, avec une attente de démonstration in planta et de consortium multi-acteurs. La phase 3 du protocole (SynComs phylogénétiquement guidés, gradient d'intensification agricole) correspond parfaitement aux attendus, mais nécessite un consortium élargi incluant des partenaires industriels et des instituts agricoles." } ], "strengths": [ "Original hypothesis combining fungal phylogeny, BGC genomics and synthetic ecology, with strong potential for generalisation beyond endophytes.", "Three-phase protocol with quantitative GO/NO-GO criteria and a modular budget (€18–120k), which facilitates alignment with low-TRL, high-risk calls.", "Positioning at the interface between biological control and community synthetic biology, a niche that is sparsely populated and attractive to European reviewers." ], "weaknesses": [ "The current TRL is very low (TRL 2–3): the in silico phase relies on data for antagonistic traits that are often heterogeneous and poorly standardised, which undermines reproducibility.", "The consortium is not defined, and the hypothesis requires a rare combination of expertise (fungal phylogenomics + plant pathology + SynComs) that is difficult to assemble within a single laboratory.", "The threshold value λ ≥ 0.3 is arbitrary and could be rejected by exacting reviewers in the absence of theoretical justification or prior power analysis." ], "critical_questions": [ "Is a sufficiently standardised phenotypic antagonism dataset (≥100 strains, ≥2 pathogens) already available to justify the in silico phase without prior collection?", "What is the plan for managing conflicts of interest and intellectual property if an industrial partner (biocontrol) joins the consortium in Phase 3?", "How is it ensured that the phylogenetic signal observed in vitro translates in planta, given that antagonistic traits are strongly modulated by the host and the soil?" ], "recommendation": "Target first a low-TRL ANR call (JCJC or PRC) to fund Phases 1–2 and produce a publishable proof of concept, then capitalise on these results to aim for an ERC Starting Grant or a Horizon Europe call (HORIZON-CL6) in Phase 3. Direct submission to the ERC without preliminary data should be avoided: the risk of rejection for lack of feasibility is too high." } ], "meta_review": { "consensus_score": 6.06, "verdict": "publish_brief", "brief_quality_gate": false, "llm_verdict": "reject", "llm_consensus_score": 6.2, "verdict_override_reason": "Python threshold override: consensus 6.06 at iter 2 → publish_brief (LLM said reject)", "key_consensus": [ "The protocol is broadly well structured, with quantitative GO/NO-GO criteria, pre-registration and a transparency approach (open data and code) that are commended by several reviewers.", "The hypothesis is conceptually original and articulates a clear causal chain between genes (BGCs), antagonistic phenotypes and in planta efficacy, which represents a potential advance for biocontrol.", "The link between in vitro antagonism and in planta efficacy is recognised as a necessary but untested condition, and the absence of prior validation weakens the prediction as a whole.", "The phylogenetic signal threshold (Pagel's λ ≥ 0.3) is judged arbitrary and unsupported by empirical data or a power analysis, which constitutes a shared weakness.", "Horizontal gene transfer (HGT) is identified as a major confounder that could invalidate the phylogenetic prediction if it is not quantified." ], "key_disagreements": [ "The methodologist and the domain expert consider that the protocol is revisable and recommend a major revision, whereas the contrarian judges that the flaws are too fundamental (FAIL REASON #1, #2, #3) to be corrected without prior validation on independent data.", "The industrialist and the funding strategist see applicative potential and funding opportunities, but emphasise that the TRL is too low and that the consortium is not defined, which contrasts with the optimism of the academic reviewers regarding feasibility.", "The domain expert insists on the need to validate the in vitro–in planta correlation on a large dataset before testing the phylogenetic signal, whereas the methodologist proposes additional controls but does not call into question the sequence of the phases.", "The contrarian asserts that the statistical power is insufficient to detect a realistic effect and that the random prediction is poorly evaluated, whereas the methodologist acknowledges the absence of a power analysis but does not judge it to be disqualifying." ], "revision_guidance": [], "critical_path": "The prior validation of the correlation between in vitro antagonism and in planta efficacy on an independent dataset, with an R² ≥ 0.5, is the most determining factor: without this evidence, the causal chain collapses and the predictions lose their agricultural relevance. In parallel, the quantification of the HGT rate of antagonistic BGCs and an a priori power analysis are indispensable for establishing the robustness of the phylogenetic signal.", "final_recommendation": "The panel recognises the originality and rigour of the protocol, but the weaknesses identified are too fundamental to be resolved within the scope of a revision. The unvalidated link between in vitro antagonism and in planta efficacy, the arbitrary threshold of λ ≥ 0.3, the absence of a power analysis and the underestimation of HGT compromise the validity of the hypothesis. The contrarian, with high confidence, notes that without prior validation on independent data, the predictions could be statistically significant in vitro but without agricultural relevance. Consequently, the panel recommends rejecting the hypothesis as it stands and inviting the authors to first produce empirical evidence of the in vitro–in planta correlation and a quantification of HGT before resubmitting." } } }