Manas Mahale

Technical Expertise

Core Competencies

Computational Chemistry Machine Learning Metabolomics

Programming & Tools

Python Git AWS RDKit Schrödinger

ML & Data Science

Machine Learning Neural Networks QSAR Chemical Language Models

Domain Knowledge

Cheminformatics Drug Discovery Molecular Modeling Pharmacokinetics Medicinal Chemistry

Education

Bombay College of Pharmacy, University of Mumbai

B.Pharmacy Mumbai Feb 2020 - 2025

Relevant Coursework

Medicinal Chemistry Biopharmaceutics & Pharmacokinetics Pharmacology Novel Drug Delivery Systems Organic Chemistry Pharmacognosy

Experience

Bender Group, Universitatea Babeş-Bolyai

Group Member Cluj

July 2024 - Present

Developing Ligand-Target Prediction and Mode-of-Action Analysis models to enhance phenotypic small molecule virtual screening
Conducting Research on Foundational Models for Single Cell Perturbation data

Pangea Botanica

AI & ML Intern Berlin

Sept 2022 - June 2023

Created LILA, a comprehensive Ligand-Target Prediction suite, streamlining the drug discovery pipeline
Improved MS2DeepScore Model performance in predicting Natural Product Structural Similarity from MS/MS Spectra
Developed and Deployed State-Of-The-Art computational prediction models for Pharmacokinetic endpoints
Established workflows for LILA and computational metabolomics module on AWS

ChemBio Discovery

Computational Chemistry Consultant Boston

June 2022 - July 2022

Automated Schrödinger workflows for the computational validation of Covalent Docking Pipelines, reducing processing time by 300%
Streamlined protein alignment and homology modeling processes using Schrödinger's Python API

Center for Molecular Informatics, University of Cambridge

Visiting Undergraduate Researcher Cambridge, UK

Feb 2022 - June 2023

Researched Tandem Mass Spectra for Molecular Similarity Prediction, focusing on the impact ofPositional Embeddings
Explored applications of Masked Language Models in medicinal chemistry for small molecular library generation

The Coutinho Lab, BCP

Undergraduate Researcher Mumbai, India

Sept 2021 - Sept 2022

Developed Chemical Language Models for Fragment Based Drug Design
Developed Analogue Split - A Chemically Biased Parametric Data Splitting Method for Model validation against IID Activity Cliff data

Volunteering

U&I, Our Lady's Home

Volunteer Mumbai, India

Sept 2023 - March 2024

Passionately contributing to the education and personal development of underprivileged children by delivering engaging and interactive lessons, and fostering a positive learning environment
Collaborating with a team of volunteers to create and implement innovative teaching methods, educational materials and raising funds to ensure holistic learning

Publications

Seal, S., Williams, D., Hosseini-Gerami, L., Mahale, M., Carpenter, A.E., Spjuth, O., Bender, A.

Improved Detection of Drug-Induced Liver Injury by Integrating Predicted In Vivo and In Vitro Data

Chem. Res. Toxicol. 37 1290–1305 (2024)

DOI: 10.1021/acs.chemrestox.4c00015

Martis, E.A.F., Mahale, M., Choudhary, A., Coutinho, E.C.

Understanding protein-ligand interactions using state-of-the-art computer simulation methods

Cheminformatics, QSAR and Machine Learning Applications for Novel Drug Development Elsevier 181–203 (2023)

DOI: 10.1016/B978-0-443-18638-7.00015-3

Gupta, N., Mohan, C.D., Shanmugam, M.K., Jung, Y.Y., Chinnathambi, A., Alharbi, S.A., Ashrafizadeh, M., Mahale, M., Bender, A., Kumar, A.P., Putti, T.C., Rangappa, K.S., Zhang, X., Ahn, K.S., Sethi, G.

CXCR4 expression is elevated in TNBC patient derived samples and Z-guggulsterone abrogates tumor progression by targeting CXCL12/CXCR4 signaling axis in preclinical breast cancer model

Environmental Research 232 116335 (2023)

DOI: 10.1016/j.envres.2023.116335

Projects

DILIPredictor

pip install dilipred

Developed DILIPredictor for in-silico prediction of drug-induced liver injury (DILI), using a combination of in vitro, in vivo, pharmacokinetic, and molecular structural data
Achieved AUC-PR of 0.79, enabling improved detection of hepatotoxic compounds and distinguishing between human and animal-specific DILI cases

Analogue Split

pip install analoguesplit

Developed tool for QSAR model robustness & validation in handling activity cliffs and evaluating model performance across different fractions of activity cliffs in test set
Created Gamma Plot visualization tools to analyze the relationship between activity cliffs and model predictions

MolecularNetwork

pip install molecularnetwork

Developed package for creating molecular similarity networks using RDKit and NetworkX, with customizable descriptors and similarity metrics
Enabled visualization and analysis of molecular networks through graph-based tools, facilitating structure-activity relationship exploration

FragmentBERT

Masked Language Models are Fragment Based Drug Designers

BERT based Masked Language Models trained on tokens of Substructure of the compound as opposed to tokenized SMILES characters to better characterize chemical space using chemically meaningful tokens
Retrospectively validated the method upon various drug classes

PKSmart

pip install pksmart

Developed package that models human pharmacokinetics (PK) parameters using molecular structural fingerprints, physicochemical properties, and predicted Animal PK data
Achieved prediction accuracies comparable to industry-standard models for key PK parameters like VDss, CL, and fu, facilitating early PK property assessment in drug discovery

Fellowship & Awards

🥇

Gold Medal & Fellowship

AIU Anveshan, 2024: Health Science - National SRC

🥈

Silver Medal

AIU Anveshan, 2024: Health Science - WEST Zone

🥈

Silver Medal & Fellowship

Avishkar Research Convention, 2024 - Maharashtra, IN

💰

Cumulative Fellowship of USD 1000

Poster & Podium Presentations FragmentBERT: Masked Language Models are Fragment Based Drug Designers

References

Prof. Dr. Andreas Bender

CTIO, Pangea Bio Ltd Professor for Molecular Informatics University of Cambridge

Prof. Dr. Evans Coutinho

Dean Research St John Technical and Education Campus