Machine Learning Identification of Functional Trait Syndromes Associated with Responsiveness to Arbuscular Mycorrhizal Fungi
- 1Mohammed VI Polytechnic University
Abstract
Arbuscular mycorrhizal fungi (AMF) are widespread plant symbionts that enhance nutrient acquisition, growth, and stress tolerance, yet plant responsiveness to AMF varies substantially and remains difficult to predict. We developed Trait2Myco, a Random Forest based machine-learning framework that integrates plant functional traits and environmental variables to predict mycorrhizal growth response (MGR). The framework was trained using 80 observations spanning four grass species grown under four precipitation regimes. Predictor variables included leaf and root economic traits together with precipitation, while total MGR served as the response variable. A full model incorporating ten functional traits and precipitation was trained using 80% of observations and evaluated on an independent testing dataset (20%, n = 16). The full model explained 41% of variation in MGR (R2 = 0.41), whereas a reduced ecological model containing only specific leaf area, specific root area, root carbon concentration, and precipitation improved predictive performance and explained 51.7% of variation (R2 = 0.517). Ten-fold cross-validation of the full model yielded an R2 of 0.983, RMSE of 0.046, and MAE of 0.036. Variable importance analysis identified specific leaf area, precipitation, root carbon concentration, and specific root area as the strongest predictors of responsiveness. Partial dependence analysis revealed nonlinear relationships between these predictors and MGR, while principal component analysis showed that highly responsive plants occupied distinct regions of functional trait space characterized by resource-acquisitive trait syndromes. Leaf and root traits contributed nearly equally to model performance, accounting for 41.9% and 41.3% of cumulative importance, respectively, while environmental factors contributed 16.8%. An exploratory analysis using an independent AMF colonization dataset produced similarly strong predictive relationships, suggesting that trait-based machine-learning approaches may generalize across multiple AMF-associated responses. Trait2Myco provides a reproducible framework for identifying plant trait syndromes associated with responsiveness to AMF and offers a practical tool for predicting mycorrhizal benefits across environmental gradients.
Related
- Beyond phosphate uptake: PT4/PT11 as key functional components linking phosphate acquisition, nutrient exchange, and arbuscule maintenance in arbuscular mycorrhizal symbiosis
- Environmental variables associated with Arbuscular Mycorrhizal Fungi diversity in Ohio soybean agricultural systems.
- Arbuscular mycorrhizal fungi and mycorrhizal helper bacteria co-inoculation alters drought-associated physiological traits and leaf abscisic acid concentration in citrus rootstocks
- Rice-straw extract directly stimulates pre-symbiotic arbuscular mycorrhizal fungal growth and is associated with greater phosphorus acquisition in Sorghum
- Soil Phosphorus Conditions Drives Arbuscular Mycorrhizal Fungal Community Differentiation and Modulate the Growth of Cenchrus pauciflorus
BibTeX
@misc{shah2026machine,
title = {{Machine Learning Identification of Functional Trait Syndromes Associated with Responsiveness to Arbuscular Mycorrhizal Fungi}},
author = {T. Shah and M. A. Errahmouni and M. Hijri},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.05.756734},
url = {https://www.biorxiv.org/content/10.64898/2026.10.05.756734v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Machine Learning Identification of Functional Trait Syndromes Associated with Responsiveness to Arbuscular Mycorrhizal Fungi
AU - T. Shah
AU - M. A. Errahmouni
AU - M. Hijri
PY - 2026
DA - 2026/10/05
PB - bioRxiv
DO - 10.64898/2026.10.05.756734
UR - https://www.biorxiv.org/content/10.64898/2026.10.05.756734v1
AB - Arbuscular mycorrhizal fungi (AMF) are widespread plant symbionts that enhance nutrient acquisition, growth, and stress tolerance, yet plant responsiveness to AMF varies substantially and remains difficult to predict. We developed Trait2Myco, a Random Forest based machine-learning framework that integrates plant functional traits and environmental variables to predict mycorrhizal growth response (MGR). The framework was trained using 80 observations spanning four grass species grown under four precipitation regimes. Predictor variables included leaf and root economic traits together with precipitation, while total MGR served as the response variable. A full model incorporating ten functional traits and precipitation was trained using 80% of observations and evaluated on an independent testing dataset (20%, n = 16). The full model explained 41% of variation in MGR (R2 = 0.41), whereas a reduced ecological model containing only specific leaf area, specific root area, root carbon concentration, and precipitation improved predictive performance and explained 51.7% of variation (R2 = 0.517). Ten-fold cross-validation of the full model yielded an R2 of 0.983, RMSE of 0.046, and MAE of 0.036. Variable importance analysis identified specific leaf area, precipitation, root carbon concentration, and specific root area as the strongest predictors of responsiveness. Partial dependence analysis revealed nonlinear relationships between these predictors and MGR, while principal component analysis showed that highly responsive plants occupied distinct regions of functional trait space characterized by resource-acquisitive trait syndromes. Leaf and root traits contributed nearly equally to model performance, accounting for 41.9% and 41.3% of cumulative importance, respectively, while environmental factors contributed 16.8%. An exploratory analysis using an independent AMF colonization dataset produced similarly strong predictive relationships, suggesting that trait-based machine-learning approaches may generalize across multiple AMF-associated responses. Trait2Myco provides a reproducible framework for identifying plant trait syndromes associated with responsiveness to AMF and offers a practical tool for predicting mycorrhizal benefits across environmental gradients.
ER -