Publications – Page 2 – SilicoScientia

Machine learning aided de novo design identifies novel Benzimidazolone based inhibitor-modulators for Heat Shock Protein 90 (HSP90)

Our research began with a machine learning-assisted de novo design process, employing REINVENT4 to design and generate novel molecular structures inspired by known HSP90 inhibitors. The generated molecules underwent a series of rigorous cheminformatics analyses to assess their pharmacokinetic properties. We then evaluated their binding affinities through docking simulations and predicted their absolute binding affinity using the KDeep tool, thereby refining the designed chemical space. From our comprehensive analyses, three benzimidazole-based drug-like candidates emerged as promising HSP90 inhibitors: IM1, IM2, and IM3. Their docking-based binding affinities were impressive, registering at -11.30, -11.50, and -11.20 kcal/mol, respectively. Published article “Machine learning aided de novo design identifies novel Benzimidazolone based inhibitor-modulators for Heat Shock Protein 90 (HSP90)” found in Chemistry Select.

Molecular Simulations and Machine Learning Methods for the Identification of Novel Aurora A Kinase Inhibitors

Aurora kinase A (AAK) is a critical regulator of mitosis, the process of cell division. It plays essential roles in cell cycle regulation, centrosome maturation, spindle assembly, and chromosome segregation, all crucial for accurate daughter cell formation. Deregulation of AAK expression and activity has been linked to various human diseases, particularly cancer, where increased AAK expression or activity contributes to cancer development and progression. In our recent study, we have employed artificial intelligence and machine learning techniques to identify potential inhibitors of AAK. Particularly we used multi-step molecular docking via AutoDock Vina and PLANTS to screen a ChemDiv kinase-specific inhibitor library against AAK. The study successfully identified three hit compounds with perfect binding in the active site pockets of AAK, comparable to the standard BindingDB compound and the co-crystal ligand VX-680 binding mode. These findings underscore the potential of computational drug discovery, bolstered by AI and machine learning, in identifying promising AAK inhibitors for improved cancer management. Published research article “Molecular Simulations and Machine Learning Methods for the Identification of Novel Aurora A Kinase Inhibitors” available in Journal of Biomolecular Structure & Dynamics.

Identification of Mycobacterium tuberculosis transcriptional repressor EthR inhibitors

Tuberculosis (TB) has plagued humanity since prehistoric times, and the rise of drug-resistant strains has only made it harder to combat. Ethionamide (ETH), a second-line drug, targets mycolic acid synthesis in TB bacteria but requires activation by the EthR protein. Our research has focused on finding new ways to enhance this activation process. In an exciting development, we have identified three promising molecules that could serve as ETH boosters by inhibiting the EthR protein. Advanced techniques like molecular docking, molecular dynamics simulations, and binding free energy studies confirmed the potential of those identified compounds as effective EthR inhibitors. Additionally, machine learning methods evaluated their toxicity and synthesisability, confirming their safety and ease of synthesis. This breakthrough offers new hope in the battle against this ancient disease. Check out this published article “Identification of Mycobacterium Tuberculosis Transcriptional Repressor EthR Inhibitors: Shape-Based Search and Machine Learning Studies” in Heliyon.

Machine learning assisted methods for the identification of low toxicity inhibitors of Enoyl-Acyl Carrier Protein Reductase (InhA)

Tuberculosis, a deadly disease with ancient roots, continues to pose a global health threat. We have employed artificial intelligence and machine learning integrated approaches to identify potential new treatments by targeting the InhA protein in the tuberculosis-causing bacteria. A large chemical database ‘SelleckChem’ was meticulously filtered using advance computational methods to pinpoint molecules with promising drug-like properties. Molecular dynamics simulations and machine learning based ADMET analyses confirmed the potential of five identified compounds to effectively bind to the InhA protein and disrupt its function. Crucially, these molecules were predicted to be safe and easily producible, offering hope for the development of new anti-tuberculosis drugs. Article “Machine Learning Assisted Methods for the Identification of Low toxicity Inhibitors of Enoyl-Acyl Carrier Protein Reductase (InhA)” can check out in Science Direct.

AI/ML Assisted De Novo Design Of Hsp90 Inhibitors

Aurora Kinase Inhibitors Identification: An AI/ML Study