AbiOmics: An End-to-End Pipeline to Train Machine Learning Models for Discrimination of Plant Abiotic Stresses Using Transcriptomic Profiling Data

Abstract Abiotic stresses are primary constraints on global crop productivity, reducing yields by up to 80%. While traditional phenotypic sensing detects stress only after physiological symptoms emerge and often fails to discriminate specific stressor types, transcriptomic profiling offers a high-dimensional solution, capturing rapid and sensitive molecular shifts. In this study, we developed AbiOmics, the first end-to-end machine learning pipeline specifically designed to identify and discriminate among multiple stressors. This approach represents a previously undocumented method for stress specification using large-scale transcriptomic big data. We identified 320 stress-specific marker genes using a curated collection of 1,243 transcriptomes of Arabidopsis samples treated with four major abiotic stresses, salt, cold, heat, and drought. A single-layer perceptron model trained on these features achieved 91% accuracy during five-fold cross-validation and 93% accuracy on an independent test set. The model demonstrated an unprecedented capacity to generalize to multi-stress conditions, identifying concurrent signatures in combinatorial salt-and-heat treatments. By integrating marker identification with SHAP-based biological interpretation, AbiOmics provides a rigorously validated diagnostic tool superior to conventional sensing. This framework establishes a high-confidence labeling strategy for AI-driven crop management and precision breeding to mitigate climate change impacts. Graphical Abstract