HT_PREDICT: a machine learning-based computational open-source tool for screening HDAC6 inhibitors.

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Histone deacetylase 6 (HDAC6) is a key therapeutic target for cancers and neurodegenerative conditions like Alzheimer's disease.
• Developing selective and non-toxic HDAC6 inhibitors is crucial for effective treatment strategies.

Purpose of the Study:

• To create robust quantitative structure-activity relationship (QSAR) models for predicting HDAC6 inhibitors.
• To integrate these QSAR models into a user-friendly web application for virtual screening.

Main Methods:

• Utilized a dataset of 3854 compounds, employing PubChem and Klekota-Roth data with 2D atom pair fingerprints and RDKit descriptors.
• Applied machine learning algorithms including gradient boosting, support vector machines, neural networks, and k-nearest neighbors.
• Developed the HT_PREDICT web application (https://htpredict.streamlit.app/) to host the QSAR models.

Main Results:

• Established validated regression QSAR models with confirmed predictive accuracy through in vitro studies.
• The HT_PREDICT application successfully performed virtual screening of compounds.
• Identified two novel, promising HDAC6 inhibitors for future preclinical investigation.

Conclusions:

• The developed QSAR models and the HT_PREDICT application provide a valuable tool for identifying potential HDAC6 inhibitors.
• This approach accelerates the drug discovery process for diseases associated with HDAC6 dysregulation.
• The identified compounds warrant further research for therapeutic development.