Chattopadhyay, Ishanu

Group Research Activities

Our research group focuses on developing advanced AI and machine learning frameworks to address complex, data-driven challenges in biomedical, epidemiological, and societal domains. A central theme is the uncovering of predictive patterns in complex systems. We have pioneered the Zero-burden Comorbid Risk (ZCoR) platform, a universal screening tool for early detection of diseases such as autism spectrum disorder (ASD), idiopathic pulmonary fibrosis (IPF), Alzheimer’s disease, and related dementias. ZCoR is optimized for sparse, noisy medical data and enables robust, population-wide risk assessments using minimal computational overhead. This work has received substantial federal support, including DARPA and NIH grants, and is being validated across multiple health systems. Our team also investigates creation of digital twins of complex biological and social systems, with applications ranging from tracking zoonotic pathogen evolution to understanding microbiome dynamics and social opinion shifts. Beyond healthcare, we also address societal modeling, such as crime prediction, using our developed Fractal-Net models, achieving high accuracy while revealing systemic biases. Our work, published in top-tier journals, has sparked interest across academia, industry, and the media, facilitating collaborations with institutions like Colorado Anschutz, Harvard Medical School, and the Veterans Health Administration. The following publications showcase the breadth and influence of our work: [https://www.nature.com/articles/s41591-022-02010-y|https://www.nature.com/articles/s41591-022-02010-y] [https://www.science.org/doi/10.1126/sciadv.adj0400|https://www.science.org/doi/10.1126/sciadv.adj0400] [https://www.nature.com/articles/s41562-022-01372-0|https://www.nature.com/articles/s41562-022-01372-0]

List all projets

Zero-burden Comorbid Risk (ZCoR) Platform Universal screening tool for complex diseases, including Autism Spectrum Disorder (ASD), Idiopathic Pulmonary Fibrosis (IPF), Alzheimer’s Disease, and Related Dementias (ADRD). Currently validated across multiple health systems, supported by DARPA, NIH, and collaborations with institutions like Mayo Clinic and Veterans Health Administration.
Digital Twins for Complex Biological and Social Systems (Q-Net) Framework to create digital twins for applications in microbiome modeling, pathogen evolution, and opinion dynamics in society. Recent applications include predicting infant gut microbiome evolution and zoonotic pathogen emergence, with ongoing work in developing a "Bio-NORAD" for pandemic prediction.
Forecasting Epidemic Spread (UnIT Score) Geospatial risk assessment tool developed for flu-like transmission patterns, validated during the COVID-19 pandemic. Focused on providing early warning signals at the county level for epidemic preparedness.
Rare and Extreme Event Prediction Using Fractal-Net Applied to crime prediction in U.S. cities, revealing biases in law enforcement. High predictive performance and collaboration with law enforcement agencies for real-world impact.
Algorithmic Lie Detector (TruthNet) Adaptive AI framework for identifying malingering in mental health diagnostics, especially in PTSD and substance use disorders. Developed in collaboration with experts in psychiatry, with prospective studies planned for incarcerated populations.
Predictive Models for Pathogen Evolution Using Q-Nets to model evolutionary pathways and predict pandemic strains and mutation trajectories in pathogens like SARS-CoV-2 and influenza. Funded by DARPA’s PREEMPT and PAI programs, with experimental validation underway.
Modeling Opinion and Belief Dynamics in Society (CogNet) Mechanistic modeling of opinion shifts and polarization in U.S. society. Future plans include large-scale human validation studies and further grants for extending these predictive societal models. 8. Developing a Digital Twin of the Human Genome for Epistasis Analysis Project to capture complex gene interactions and environmental influences in diseases with intricate genomic footprints, such as interstitial lung disease. Collaborative effort leveraging biobank data to explore epistasis and gene-environment interactions.

Computational methods

Our computational methods comprise implementations of advanced machine learning algorithms, most of which are developed within our group. We also validate with standard ML tools including tensorflow and other learning suites in Python. We also use compiled C++ codes widely where fast execution and custom architectures are warranted.

List all Software

Python, Python packages such as Scipy, Scikitlearn and related libraries, R, C++ compiler, Paraview, Boost Libraries.

UKY Collaborators

Mark Ebbert, Jefferey Talbert, Darwin Conwell

External collaborators: Robert Gibbons, Professor, University of Chicago; James Evans, Professor, University of Chicago; David Schwartz, Professor, Colorado; Anschutz Fernenado Martinez, Professor, UMass;