Machine Learning-Based Prediction of Abiotic Stress-Responsive Genes in Date Palm (Phoenix dactylifera L.)

Abiotic stresses pose major constraints to Phoenix dactylifera L. productivity in different environments. Although date palm possesses substantial adaptive capacity, experimentally validated stress-responsive gene resources remain limited. Here, we developed a sequence-based machine learning framework to prioritize candidate abiotic stress-responsive genes in date palm. Training labels were derived computationally: the 48,801-protein reference proteome was annotated by best-hit orthology to Arabidopsis thaliana, and stress-associated Gene Ontology terms transferred from these orthologs were used to assign cold, heat, drought, and salinity labels. Sequence-derived features were evaluated with multiple classifiers, followed by SVM-based recursive feature elimination and hyperparameter optimization. The final stress-specific models achieved held-out test auROC values of 80.14% for cold, 85.81% for heat, 83.76% for drought, and 84.28% for salinity. Under leakage-aware nested cross-validation that kept proteins sharing a gene or at least 40% sequence identity in the same partition, however, these k-mer models reached auROC values of only 51.9–69.2%, indicating that much of their apparent accuracy reflected similarity to training proteins, whereas models built on ESM-2 protein language model embeddings retained auROC values of 69.1–74.9%. A revised multi-label formulation retained genes associated with multiple stresses and used a genome-wide background class of putative negatives, yielding 426 cold-, 261 heat-, 559 drought- and 699 salt-associated candidate genes at the SVM decision boundary. The multi-task model additionally predicted drought–salinity genes as the most prominent combination. Overall, the study provides a genome-scale, hypothesis-generating prioritization resource for date palm abiotic stress research.

Authors

Institutions

Publication Details

Journal
Plants
Published
2026-11-18
DOI
https://doi.org/10.3390/plants15223617
Primary Topic
Machine Learning in Bioinformatics
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Machine Learning-Based Prediction of Abiotic Stress-Responsive Genes in Date Palm (Phoenix dactylifera L.)

Hidayah M Alotaibi, Muhammad Haroon, Mahnoor Basit, Rola Alhamad
Plants
Machine Learning in Bioinformatics
article

Machine Learning-Based Prediction of Abiotic Stress-Responsive Genes in Date Palm (Phoenix dactylifera L.)

Hidayah M Alotaibi, Muhammad Haroon, Mahnoor Basit, Rola Alhamad
article en

Abstract

Abiotic stresses pose major constraints to Phoenix dactylifera L. productivity in different environments. Although date palm possesses substantial adaptive capacity, experimentally validated stress-responsive gene resources remain limited. Here, we developed a sequence-based machine learning framework to prioritize candidate abiotic stress-responsive genes in date palm. Training labels were derived computationally: the 48,801-protein reference proteome was annotated by best-hit orthology to Arabidopsis thaliana, and stress-associated Gene Ontology terms transferred from these orthologs were used to assign cold, heat, drought, and salinity labels. Sequence-derived features were evaluated with multiple classifiers, followed by SVM-based recursive feature elimination and hyperparameter optimization. The final stress-specific models achieved held-out test auROC values of 80.14% for cold, 85.81% for heat, 83.76% for drought, and 84.28% for salinity. Under leakage-aware nested cross-validation that kept proteins sharing a gene or at least 40% sequence identity in the same partition, however, these k-mer models reached auROC values of only 51.9–69.2%, indicating that much of their apparent accuracy reflected similarity to training proteins, whereas models built on ESM-2 protein language model embeddings retained auROC values of 69.1–74.9%. A revised multi-label formulation retained genes associated with multiple stresses and used a genome-wide background class of putative negatives, yielding 426 cold-, 261 heat-, 559 drought- and 699 salt-associated candidate genes at the SVM decision boundary. The multi-task model additionally predicted drought–salinity genes as the most prominent combination. Overall, the study provides a genome-scale, hypothesis-generating prioritization resource for date palm abiotic stress research.

PlantsVol. 15(22)
Princess Nourah bint Abdulrahman University (SA), National University of Sciences and Technology (PK)
Zero hunger
Machine Learning in Bioinformatics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon