Dr Sahanawaz

Independent Academic Research Laboratory in Chemical Data Science & Computational Chemistry

Bayesian Chemistry Lab

Molecular Machine Learning & Chemical Data Science

Core Identity: Prior chemical principles updated by experimental observation under quantified uncertainty.

Department of Science & Humanities, Sheikhpara ARM Polytechnic, Govt. of West Bengal
Doctoral Lineage: IISER Kolkata (PhD, 2026) · IIT Bombay (M.Sc., 2018)

We approach chemical machine learning through a Bayesian perspective: treating established physical-organic principles as informative priors, updating model predictions as new experimental evidence arrives, and systematically quantifying uncertainty and applicability domain boundaries in molecular property prediction and chemical sensing.

Dr. Md Sahanawaz
Lead Researcher

Dr. Md Sahanawaz, PhD (IISER Kolkata) · M.Sc. (IIT Bombay)

Lecturer & Head, Department of Science & Humanities

Sheikhpara ARM Polytechnic, West Bengal, India

Dr. Md Sahanawaz is a chemist and computational researcher. He completed his B.Sc. in Chemistry from Ramakrishna Mission Vivekananda Centenary College (2016), M.Sc. in Organic Chemistry from IIT Bombay (2018), and PhD in Chemistry from IISER Kolkata (2026) under the supervision of Prof. Subhajit Bandyopadhyay. He qualified CSIR-NET with All India Rank 17 and was selected through the West Bengal Public Service Commission (WBPSC). Alongside his teaching responsibilities at Sheikhpara ARM Polytechnic, he conducts independent research focused on molecular machine learning, cheminformatics, explainable AI, and spectroscopic pattern recognition.

CSIR-NET (AIR 17) GATE JAM WBPSC Selection
iD ORCID: 0009-0000-6615-5441 GS Google Scholar GH GitHub in LinkedIn ✉ sahanawaz@wbscte.ac.in
Research Areas

Current Research Focus

Our computational investigations span rigorous model evaluation, interpretability, uncertainty quantification, and predictive modeling for chemical systems.

01

Critical Evaluation of ML Models in Chemistry

Machine learning models in chemistry frequently report over-optimistic performance due to naive random splitting, hidden data leakage, and memorization of dataset artifacts. We investigate rigorous validation methodologies, including Bemis–Murcko scaffold splits, temporal splits, and cluster-based validation, benchmarking models against simple chemical baselines to ensure genuine predictive capability.

  • Scaffold Validation
  • Data Leakage Diagnostics
  • Baseline Benchmarking
  • Cross-Validation Protocols
  • Model Generalizability
02

Explainable AI (XAI) & Chemical Interpretability

Black-box predictions are inadequate for scientific discovery where mechanistic understanding is paramount. We develop and apply feature-attribution and substructure-level explanation techniques (such as SHAP, Integrated Gradients, and molecular attribution maps) to interrogate what chemical features drive predictions, verifying whether learned representations align with physical organic principles.

  • Feature Attribution
  • Substructure Importance
  • Model Interpretability
  • Mechanistic Plausibility
  • Counterfactual Analysis
03

Uncertainty Quantification & Applicability Domain Estimation

For computational predictions to be useful to experimentalists, models must know when they do not know. We implement applicability domain (AD) boundaries and uncertainty quantification (UQ) methods—using distance-to-model metrics, ensemble variance, and conformal prediction—to establish calibrated confidence intervals for every single molecular query.

  • Applicability Domain (AD)
  • Conformal Prediction
  • Ensemble Uncertainty
  • Reliability Bounds
  • Chemical Space Distance
04

High-Quality Chemical Data Curation

The reliability of chemical ML is fundamentally constrained by data quality. We build automated data-curation workflows for chemical datasets: canonical SMILES standardization, salt stripping, stereochemistry normalization, duplicate reconciliation, and systematic detection and handling of missing or noisy measurements in experimental and spectral datasets.

  • SMILES Standardization
  • Curated Benchmarks
  • Missing Data Imputation
  • Spectral Cleaning
  • Noise Detection
05

Chemistry-Aware Machine Learning Models

Standard statistical algorithms often fail to respect fundamental chemical invariants. We work on incorporating domain constraints—such as molecular graph symmetry, charge neutrality, thermodynamic consistency, and physical group additivity—directly into model architectures and loss formulations as chemical priors to improve sample efficiency and generalization.

  • Chemical Priors
  • Physics-Informed ML
  • Symmetry & Invariants
  • Graph Architectures
  • Thermodynamic Bounds
06

Property Prediction of Photochromic Molecules

Photoresponsive molecular systems (e.g., azobenzenes, photoswitches) exhibit complex, state-dependent behavior. We build targeted predictive models for photochromic properties, including thermal cis–trans isomerization rates, activation energy barriers, absorption maxima, and half-lives across different solvent environments and polymer architectures.

  • Azobenzene Photoswitches
  • Isomerization Kinetics
  • Thermal Half-Lives
  • State-Dependent Properties
  • Polymer Microenvironments
07

Chemical Sensing & Multivariate Pattern Recognition

Building on doctoral research at IISER Kolkata, we couple molecular sensor arrays with multivariate statistical methods (PCA, LDA, clustering) to extract meaningful discrimination signatures from complex, multidimensional spectroscopic data, enabling robust identification of closely related analytes such as nucleotide phosphates.

  • Optical Sensor Arrays
  • Principal Component Analysis (PCA)
  • Linear Discriminant Analysis (LDA)
  • Spectroscopic Decomposition
  • Multi-Analyte Discrimination
Doctoral Research 2018 – 2026

IISER Kolkata

Supervisor: Prof. Subhajit Bandyopadhyay

Photochromic Receptors for Nucleotide Detection Using Machine Learning Techniques

PhD in Chemical Sciences

Doctoral research investigated photoswitchable tripodal azobenzene receptors capable of differential binding with nucleotide phosphates (ATP, ADP, GTP, UMP, UTP, CTP) and inorganic phosphates. By tracking UV–Vis absorption spectra across both photoisomeric states and applying multivariate statistical analysis (PCA and LDA), the work demonstrated clear analyte discrimination based on optical response patterns, providing the foundation for our current work in data-driven chemistry.

Methodology

Computational Tools & Environment

We prioritize open-source, reproducible Python workflows for chemical data processing and modeling.

Core Environment

Python 3.10+ scientific ecosystem, NumPy, SciPy, pandas, Jupyter, virtualized reproducible runtimes.

Cheminformatics

RDKit, molecular graphs, 2D/3D conformers, Morgan/circular fingerprints, MACCS keys, physicochemical descriptors.

Machine Learning & XAI

scikit-learn, Random Forests, Gradient Boosting, Gaussian Process Regression, SHAP, feature attribution.

Spectroscopy & Data Analysis

UV–Vis absorption time-series, TCSPC multi-exponential reconvolution, multi-format instrument parsers (Horiba, JASCO, Shimadzu).

Software

Web-Based Tools & Code

Open-access utilities and repositories developed to support laboratory spectroscopists and chemical data analysis.

Web Application

Spectra Autopilot

UV–Vis Spectrophotometry Plotter & File Parser

An open web utility for experimental spectroscopists. Upload raw spectrophotometer exports (TXT, CSV, TSV), preview spectral curves, overlay multiple concentrations, customize axes, and export publication-ready vector and raster plots without coding.

Stack: Python · Streamlit · Plotly
Open Resource →
Web Application

TCSPC AutoFit

Fluorescence Lifetime Decay Fitting via IRF Reconvolution

A dedicated browser tool for time-correlated single photon counting (TCSPC) data. Parses instrument export files and performs iterative reconvolution of the Instrument Response Function (IRF) with multi-exponential decay models to extract fluorescence lifetimes.

Stack: Python · Streamlit · SciPy Optimize
Open Resource →
Open Source

Cheminformatics & ML Repositories

Open-Source Python Workflows

Scripts and reproducible notebooks for chemical dataset curation, molecular descriptor generation, scaffold-based model validation, and spectroscopic data analysis.

Stack: Python · RDKit · scikit-learn · GitHub
Open Resource →
Publications

Peer-Reviewed Journal Articles

Scholarly articles published in peer-reviewed international chemistry journals.

2026
Analysis & Sensing · Vol. 6 (1), e202500082

A Smart Photoswitchable Sensor for Differential Detection of Multiple Nucleotides In Two Photoswitchable States using Machine Learning Techniques

Authors: Md Sahanawaz*, Manik Lal Maity, Sudeep Koppayithodi, Subhajit Bandyopadhyay*

Demonstrated a photoswitchable tripodal receptor undergoing reversible cis–trans isomerization with differential binding affinities toward ATP, ADP, GTP, UMP, UTP, CTP, and inorganic phosphates. Differential UV–Vis spectral responses coupled with PCA and LDA enable robust multidimensional discrimination of nucleotides.

Molecular Machine Learning Chemical Sensing Pattern Recognition (PCA/LDA) Photochromic Receptors
DOI: 10.1002/anse.202500082 ↗
2024
Journal of Physical Organic Chemistry · Vol. 37 (4), e4599

Sequence effects on the thermal cis–trans isomerization of side-chain stearate-containing azobenzene polymers

Authors: Md Sahanawaz, Manik Lal Maity, Krishna Gopal Goswami, Pintu Sar, Priyadarsi De, Subhajit Bandyopadhyay*

Synthesized block and random copolymers containing stearic acid and azobenzene side-chains via RAFT polymerization. Revealed how polymer sequence and local microenvironmental polarity modulate transition-state stabilization and isomerization kinetics.

Physical Organic Chemistry Isomerization Kinetics Polymer Chemistry Spectroscopic Modeling
DOI: 10.1002/poc.4599 ↗
2021
ACS Applied Electronic Materials · Vol. 3 (1), 309–315

Photoswitchable Molecular Glue for Carbon Nanotubes Reversibly Controls Electronic Mobility with Light

Authors: Monochura Saha, Vishal G. More, Mahmood D. Aljabri, Sheelbhadra Chatterjee, Md Sahanawaz, Subhajit Bandyopadhyay*, Sheshanath V. Bhosale*

Developed a photoresponsive azobenzene-based supramolecular molecular glue that non-covalently functionalizes carbon nanotubes, achieving reversible light-triggered electronic mobility control.

Supramolecular Chemistry Carbon Nanotubes Molecular Electronics Photofunctional Materials
DOI: 10.1021/acsaelm.0c00857 ↗

Academic Citation Profiles

Complete publication records and citation metrics are available on external scholarly databases.

Google Scholar Profile ↗ ORCID Record ↗ GitHub Profile ↗
Collaboration

Research Inquiries & Collaborations

Open to academic collaboration and student project mentorship.

Our computational research is currently self-supported and conducted independently alongside academic teaching responsibilities. We welcome collaborative discussions with experimental chemistry groups (photochemistry, supramolecular chemistry, sensor development, and synthetic chemistry) who are interested in joint computational modeling, multivariate data analysis, or developing joint grant proposals.

We also offer research project mentorship for motivated diploma and undergraduate students in chemistry, chemical engineering, and computer science who wish to gain hands-on experience in Python, RDKit, and molecular data science.

Institutional Email: sahanawaz@wbscte.ac.in  ·  Research Email: drsahanawazofficial@gmail.com

Get in Touch → Download Curriculum Vitae (PDF)
Chronology

Academic Milestones

  • 2026

    PhD Degree Awarded

    Completed and defended doctoral thesis 'Photochromic Receptors for Nucleotide Detection Using Machine Learning Techniques' at the Department of Chemical Sciences, IISER Kolkata.

  • 2026

    Publication in Analysis & Sensing

    Published research on machine-learning-assisted nucleotide detection using photoswitchable tripodal receptors (Analysis & Sensing, DOI: 10.1002/anse.202500082).

  • 2025

    Release of Web-Based Scientific Tools

    Deployed Spectra Autopilot and TCSPC AutoFit on Streamlit Cloud to provide free, accessible data processing for experimental spectroscopists.

  • 2024

    Publication in Journal of Physical Organic Chemistry

    Published work on sequence effects and local polarity on the thermal cis–trans isomerization kinetics of azobenzene copolymers (DOI: 10.1002/poc.4599).

  • 2021

    Appointed Lecturer in Chemistry

    Joined Sheikhpara ARM Polytechnic through West Bengal Public Service Commission (WBPSC) selection; currently serving as Head, Department of Science & Humanities.