LSAP: A Machine Learning Method for Leaf-Senescence-Associated Genes Prediction

Plant leaves, which convert light energy into chemical energy, serve as a major food source on Earth. The decrease in crop yield and quality is caused by plant leaf premature senescence. It is important to detect senescence-associated genes. In this study, we collected 5853 genes from a leaf senescence database and developed a leaf-senescence-associated genes (SAGs) prediction model using the support vector machine (SVM) and XGBoost algorithms. This is the first computational approach for predicting SAGs with the sequence dataset. The SVM-PCA-Kmer-PC-PseAAC model achieved the best performance (F1score = 0.866, accuracy = 0.862 and receiver operating characteristic = 0.922), and based on this model, we developed a SAGs prediction tool called "SAGs_Anno". We identified a total of 1,398,277 SAGs from 3,165,746 gene sequences from 83 species, including 12 lower plants and 71 higher plants. Interestingly, leafy species showed a higher percentage of SAGs, while leafless species showed a lower percentage of SAGs. Finally, we constructed the Leaf SAGs Annotation Platform using these available datasets and the SAGs_Anno tool, which helps users to easily predict, download, and search for plant leaf SAGs of all species. Our study will provide rich resources for plant leaf-senescence-associated genes research.