Deep learning-driven QSAR and micro-scale MD simulation-guided strategy reveals non-toxic human HGFR inhibitors

Area of Science:

• Oncology
• Computational Chemistry
• Drug Discovery

Background:

• Hepatocyte Growth Factor Receptor (HGFR) overexpression drives ovarian cancer progression.
• Existing HGFR inhibitors have limitations in potency and toxicity.

Purpose of the Study:

• To develop a computational framework for identifying novel, safe HGFR inhibitors for ovarian cancer.
• To repurpose FDA-approved drugs as potential HGFR-targeting therapeutics.

Main Methods:

• Integrated in silico approach: deep learning bioactivity prediction, structure-based drug repurposing, and D-MPNN toxicity profiling.
• ANN model trained on HGFR bioactives to screen 1,040 FDA-approved drugs.
• Molecular docking, molecular dynamics simulations, and D-MPNN toxicity assessment.

Main Results:

• Identified venetoclax, LSM-5313, and cefoperazone as lead HGFR inhibitors.
• Confirmed stable and favorable binding of lead compounds via extensive simulations and analyses.
• D-MPNN toxicity assessment indicated no significant toxic liabilities for the identified drugs.

Conclusions:

• The computational framework successfully identified promising, non-toxic HGFR inhibitors.
• Repurposed FDA-approved drugs offer viable therapeutic options for ovarian cancer.
• Further preclinical and clinical studies are warranted for the identified candidates.