Fragment-based drug design coupled with AI/ML prediction enables identification of novel PARP-1 inhibitors against triple-negative breast cancer.

Triple-negative breast cancer occurs as a formidable challenge in cancer research due to the characteristically aggressive behaviour, poor prognosis, and scarce availability of treatment options. The approval of PARP-1 inhibitors against Triple-negative Breast Cancer (TNBC) paves the way for drug discovery endeavours in this arena. The research work aimed to identify PARP-1 targeting novel inhibitors by amalgamated application of Fragment-based, accessory structure-based, and AI/ML-based techniques. Fragment hybridization and subsequent docking-based virtual screening of the compounds obtained after fragment elaboration resulted in 71 hits with good binding characteristics to PARP-1. Further, the in silico pharmacokinetic scrutiny eliminated the non-ADME compliant compounds, resulting in twenty-two hits with favourable properties. The machine learning based prediction of PARP-1 inhibitory potential followed, resulting in eight prospective PARP-1 inhibitors with structural similarity to known inhibitors. Also, AI-based assessment of anti-cancer potential laid the foundation for their prospective role against TNBC cells. Three best hits were chosen as per the predicted anti-cancer potential using MDA-MB-231, MDA-MB-436, and MDA-MB-468 cells. These hits, along with the reference ligand Olaparib, were further examined for stability in the PARP-1 cavity under physiological conditions, using molecular dynamics simulations of 500 ns. Stable trajectory profiles for most of the simulation course were observed, further confirming the potential of the chosen hits as PARP-1 inhibitors against TNBC. Synthetic accessibility predictions provided a positive future direction for the progression of compounds in the drug discovery pipeline. Further, a representative framework for the design of new PARP-1 inhibitors emerged based on findings of the study.