Identifying key traits for heat stress tolerance in wheat using machine learning

This study aimed to investigate the effectiveness of machine learning techniques in identifying and prioritizing key traits associated with heat stress tolerance in wheat. Two datasets comprising 203 and 236 wheat genotypes, previously evaluated under normal and heat stress conditions, were analyzed. Machine learning algorithms, including k-Nearest Neighbors (KNN), Support Vector Machines (SVM), Random Forest (RF), and Artificial Neural Networks (ANN), were employed to model the relationships between traits and grain weight of five spikes under heat stress. Results indicated that SVM and ANN models exhibited superior performance in predicting the target trait, with R-squared values approaching 1.0. Correlation analysis and dendrogram analysis highlighted distinct patterns in trait relationships under normal and stress conditions, emphasizing the importance of considering environmental context when studying trait interactions. The analysis of feature importance consistently revealed traits such as the number of grains per spike, days to heading, and 100-grain weight as key characteristics, repeatedly highlighted across different algorithmic approaches, underscoring their fundamental role in heat stress tolerance. The identified key traits can serve as potential targets for genetic manipulation or selection, contributing to the development of heat-tolerant wheat cultivars. The findings of this study highlight the efficacy of machine learning in expediting the breeding of heat-tolerant wheat cultivars.

Identifying key traits for heat stress tolerance in wheat using machine learning Keywords:

Artificial Neural Network (ANN) , Precision breeding , Predictive Modeling , Support Vector Machines (SVM) , Trait Importance