NRPreTo: A Machine Learning-Based Nuclear Receptor and Subfamily Prediction Tool

The nuclear receptor (NR) superfamily includes phylogeneticallyrelated ligand-activated proteins, which play a key role in variouscellular activities. NR proteins are subdivided into seven subfamiliesbased on their function, mechanism, and nature of the interactingligand. Developing robust tools to identify NR could give insightsinto their functional relationships and involvement in disease pathways.Existing NR prediction tools only use a few types of sequence-basedfeatures and are tested on relatively similar independent datasets;thus, they may suffer from overfitting when extended to new generaof sequences. To address this problem, we developed Nuclear ReceptorPrediction Tool (NRPreTo), a two-level NR prediction tool with a uniquetraining approach where in addition to the sequence-based featuresused by existing NR prediction tools, six additional feature groupsdepicting various physiochemical, structural, and evolutionary featuresof proteins were utilized. The first level of NRPreTo allows for thesuccessful prediction of a query protein as NR or non-NR and furthersubclassifies the protein into one of the seven NR subfamilies inthe second level. We developed Random Forest classifiers to test onbenchmark datasets, as well as the entire human protein datasets fromRefSeq and Human Protein Reference Database (HPRD). We observed thatusing additional feature groups improved the performance. We alsoobserved that NRPreTo achieved high performance on the external datasetsand predicted 59 novel NRs in the human proteome. The source codeof NRPreTo is publicly available at https://github.com/bozdaglab/NRPreTo.