Predicting mTOR inhibitors with a classifier using recursive partitioning and Naïve Bayesian approaches.

Area of Science:

• Computational chemistry
• Drug discovery
• Bioinformatics

Background:

• Mammalian target of rapamycin (mTOR) regulates critical cellular processes like growth and metabolism.
• Developing clinical drugs targeting mTOR is a significant area of pharmaceutical research.

Purpose of the Study:

• To develop in silico models for predicting mTOR inhibitors and non-inhibitors.
• To create a web server for online compound prediction and virtual screening.

Main Methods:

• Collected and categorized 1,264 compounds as mTOR inhibitors or non-inhibitors.
• Utilized recursive partitioning (RP) and naïve Bayesian (NB) methods with physicochemical descriptors, fingerprints, and atom center fragments (ACFs).
• Constructed 253 classification models.

Main Results:

• Achieved predictive accuracies over 90% on both training and external test sets.
• Demonstrated scaffold hopping ability by predicting 37 new mTOR inhibitors.
• Developed a web server (http://rcdd.sysu.edu.cn/mtor/) based on ACFs and Bayesian classification.

Conclusions:

• Successfully developed in silico models for predicting mTOR inhibitors using RP and NB methods.
• The "mTOR Predictor" web server facilitates compound prediction and virtual screening.
• Identified favorable and unfavorable fragments for mTOR inhibitors, aiding lead optimization and new drug design.