An Ensemble Method for Predicting and Designing of Druggable Proteins

Over the years, several proteins and peptides with diverse therapeutic properties such as anticancer, antimicrobial, antihypertensive effects have been discovered. However, only a few hundred proteins are considered druggable with US FDA approval, while most of them failed during clinical trials.

This study systematically investigates the compositional and physicochemical properties of FDA-approved proteins to develop predictive models for identifying druggable proteins. Our main dataset comprises of 356 FDA approved proteins as positive dataset and equal number of randomly selected proteins as negative dataset. We used 80% data as training set and 20% as independent validation data, with no protein in validation dataset having more than 40% similarity with any protein in training dataset.

Random forest-based model developed using SVC-L1 selected features obtained maximum performance AUC of 0.80 with MCC 0.61 on validation data. In addition to this, we performed MERCI-based motif analysis to find exclusive motifs/ patterns in druggable proteins. Finally, we proposed an ensemble-based method combining best performing machine learning model with exclusive motifs and achieved AUC 0.92 with MCC 0.83 on independent validation dataset.

In order to serve the scientific community, web server and standalone package of "ThPPred" facilitating prediction and designing of druggable proteins is proposed (https://webs.iiitd.edu.in/raghava/thppred/).