Neural Network, Random Forest Classifier and Monte Carlo Sampling Based Binding Affinity and Pharmacological Analysis of Natural Analogues of Raf-1 Kinase Inhibitor for Breast Cancer Therapeutics

Raf-1 kinase activity in the MAPK/ERK signalling pathway has been linked to increased proliferation in breast cancer. The existing drugs have shown limited therapeutic efficacy. Thus, the study aimed to identify ligands targeting Raf-1 through artificial intelligence approaches, coupled with Monte-Carlo sampling docking algorithms. The study reviews existing sampling algorithms, along with the application of neural network and random forest classifier-based algorithms with under-the-hood molecular docking. A systematic database search for natural analogues of Sorafenib, the leading Raf-1 drug, was carried out from ZINC (>37 billion compounds) and SuperNatural 3.0 (445000 compounds). NNScore 2.0 (neural network), and DeepChem (random forest classifier) predicted binding scores of natural ligands against Raf1 Kinase Protein. Glide (in Schrodinger Suite 2023-1) which runs on Monte Carlo sampling, was used for molecular docking. MMGBSA analysis (OPLS_2005 force field and PRIME VSGB 2.0 solvation model) predicted free binding energies of ligand- Raf-1 complexes. Pharmacological (ADME/Tox) properties of the ligands were predicted to evaluate their safety profile. The natural ligand, ZINC000002148905 (binding energy= -56.66 kcal/mol), stood as the best natural analogue among the top four ligands from all the programs. In this artificial intelligence- driven pharmacoinformatic analysis, the natural analogues were modelled for further in vitro experimental analysis to inhibit Raf-1 and pave the way for natural therapeutic options for the treatment of breast cancer.