Multiple Machine Learning Based-Chemoinformatics Models for Identification of Histone Acetyl Transferase Inhibitors

Area of Science:

• Biochemistry
• Medicinal Chemistry
• Computational Biology

Background:

• Histone acetyl transferases (HATs) regulate gene transcription through histone acetylation.
• GCN5 HAT is implicated in cancer progression, yet few inhibitors are known.
• Targeting GCN5 HAT offers a potential strategy for cancer therapy.

Purpose of the Study:

• To identify novel GCN5 HAT inhibitors using in-silico virtual screening.
• To evaluate the anti-cancer potential of identified compounds.

Main Methods:

• Virtual screening of the Maybridge library using machine learning models (k-NN, Naïve Bayesian, Random Forest, SVM).
• Predictive chemoinformatics model development and validation.
• Molecular docking of consensus hits into the GCN5 HAT active site.
• In vitro biological evaluation of selected molecules for HAT inhibition and anti-cancer activity.

Main Results:

• Random Forest and SVM models demonstrated comparable performance for virtual screening.
• Ten candidate molecules were selected based on virtual screening and docking.
• Three out of ten molecules exhibited significant in vitro HAT inhibition.
• Selected compounds showed weak inhibition of MCF-7 cancer cell growth.

Conclusions:

• Computational approaches successfully identified potential GCN5 HAT inhibitors.
• The identified compounds show promise as starting points for developing novel anti-cancer agents.
• Further structural optimization is recommended to enhance anti-cancer efficacy.