Exploring Molecular Descriptors and Fingerprints to Predict mTOR Kinase Inhibitors using Machine Learning Techniques.

Area of Science:

• Biochemistry
• Molecular Biology
• Pharmacology

Background:

• Mammalian Target of Rapamycin (mTOR) is a critical kinase in cancer's autophagy pathway.
• Autophagy's dual role in tumor progression complicates therapeutic targeting of mTOR.
• mTOR functions via two complexes, mTORC1 and mTORC2, targeted by kinase inhibitors.

Purpose of the Study:

• To develop predictive models for identifying mTOR kinase inhibitors.
• To explore molecular descriptors and fingerprints for feature extraction.
• To integrate traditional and deep learning approaches for enhanced prediction.

Main Methods:

• Utilized machine learning techniques for predictive model development.
• Employed Random Forest for descriptor importance and autoencoders for fingerprint identification.
• Built and validated models using identified molecular features and their combinations.

Main Results:

• Identified twenty best-performing molecular descriptors for predicting mTOR kinase inhibitors.
• Selected the optimal model based on Mathew correlation coefficient for further screening.
• Successfully integrated traditional and deep learning for feature extraction in mTOR inhibitor prediction.

Conclusions:

• This study presents a novel approach for predicting mTOR kinase inhibitors.
• The findings contribute to understanding molecular features crucial for targeting the mTOR pathway in cancer.
• The integrated methodology offers a powerful tool for accelerating drug discovery for mTOR-related diseases.