SCCGs_Prediction: a machine learning tool for prediction of sulfur-containing compound associated genes

Abstract

Sulfur-containing compounds (SCCs) are pivotal secondary metabolites widely distributed in plants, particularly within the Brassicaceae family. These compounds play crucial roles in human health and in interactions between plants and pests. In this groundbreaking study, we harnessed the extensive SuCComBase database, harvesting 1,285 protein sequences associated with sulfur-containing compounds. Employing the SVM algorithm, we pioneered the development of a predictive model for plant SCCGs, representing a novel computational approach based on sequence data. Remarkably, our SVM-Kmer model delivered exceptional performance metrics (F1score = 0.945, ACC = 0.938, AUC = 0.936). Building upon this achievement, we introduced the SCCGs_Prediction tool, a resource born of our model. Through this tool, we identified an astounding 51,638 SCCGs from a staggering 2,873,697 protein sequences spanning 81 different species. Intriguingly, our findings highlighted that the Brassicaceae and Papilionoideae subfamilies exhibit a notably higher prevalence of SCCGs compared to other plant families. In our commitment to facilitate enhanced utilization of the SCCGs_Prediction tool and the extensive plant SCCGs datasets, we have established the Sulfur-Containing Compounds Platform (SCCP). We firmly believe that the SCCP will serve as an invaluable resource hub, providing comprehensive information to the SCCs research community.