Vivek Kanpa

Precision -Omics Lab, Mount Sinai

Goal: To train deep learning algorithms on genomic biobanks to identify rare genetic variants associated with disease.

AI_PREMie, UCD Conway Institute of Biomolecular Research

Goal: To develop a best-in-class AI-assisted clinical decision making tool for maternal preeclampsia.

Preeclampsia (PE) is a maternal hypertensive disorder that claims the lives of 50,000 mothers and 500,000 babies annually. Because its symptoms are very generic, early and accurate detection of PE is a difficult task.

Outcomes: I helped develop a best-in-class AI diagnostic aid and maternal risk stratification tool for PE to assist clinicians tailor patient care strategies. I improved model sensitivity and recall (PPV) by >10% for PE status prediction and developed and evaluated a predicted PE severity score, which demonstrated strong correlation with physician-recorded severity status (no/low/high severity). Feature analysis revealed insight into potentially novel, high-specificity biomarkers for PE diagnosis. I will present an excerpt of this at the BMEII Symposium in New York.

Revolution Medicines

Goal: perform hierarchical clustering on AlphaFold-predicted structures to inform cancer target strategic priorities.

In cancer pharmacology, there is great interest in identifying common actionable features between oncogenic structures to reduce wastefulness of lead discovery.

Outcome: I scraped 600+ cancer driver proteins and hierarchically clustered them by structural similarity. I then rendered a polar dendrogram where nodes represented oncoproteins and encoded patient mutation frequency (AACR GENIE) in the node's radius. I also developed a tool to swap GENIE data with any other cancer datasets of interest (cBioPortal, CRISPRko screens, COSMIC). I documented this project's workflow over here.

Goals: develop generalizable tools to increase pharmacogenomics throughput and expidite ADME screening.

• Trained graph convolution network (DeepChem) using molecular featurization of RevMed hit library to improve permeability (Caco-2) predictions by 28% compared to conventional molecular weight cutoff.
• Automated mass spectrometry (MS) cross-linking data analysis, calculating a normalized percent cross-linking value given raw relative abundance outputs. I enabled high throughput screening for drug-target binding, which shaved dozens of hours off each 384-well MS experiment.
• Performed in vitro dose-response assays of KRAS-mutant CRC cell lines and quantified c-Myc mRNA (PAGE, qRT-PCR) and protein (BCA assay, western blotting/ELISA) expression to understand downstream effect of KRAS inhibition on Myc activity.

Apfeld Lab, Northeastern University

Goal: determine the mechanism of action for the TGF-B/smad peroxide resistance response in C. elegans nematodes.

The peroxide immune response is key to managing ROS load in C. elegans neurons. The TGF-B pathway mediates this response and is tightly regulated, leaving the peroxide defense mechanism of action unknown.

Outcome: I performed lifespan and behavioral assays on type I and type II TGF-B ligand/receptor RNAi nematode knockouts during peroxide exposure. I observed a signal integration phenotype between type I and type II TGF-B that causes half the double ligand knockouts to live longer than wildtype. I followed this with bulk RNA-seq and GO analysis of C. elegans smad knockouts (TGF-B receptor-binding proteins) to trace differential gene expression from TGF-B silencing. I discovered that both smad-2 and smad-4 knockouts overexpressed lon-1, which is found abundantly in peroxide-sensing neurons and implicated in negative regulation of the TGF-B pathway.

Takeda Pharmaceuticals

Goal: develop bioanalytic software systems to increase throughput for immunotherapeutic discovery.

• Designed end-to-end Django web apps for data processing of protein quantification assay (HTP Lunatic) and HPLC chromatography (Agilent 1260 Infinity), increasing analysis throughput, data availability, and data standardization across multi-national team.
• Built an assay fulfillment translator to operationalize the Lynx liquid handler by interpreting a scientists' planned experimental plate layout (96/384-well). Increased assay fulfillment throughput exponentially with picomolar precision.

* asterisk indicates co-first author.

David D'Onofrio*, Shrey Patel*, Nandini Samanta*, Mikio Tada*, Vivek Kanpa*. Cost-benefit analysis of non-AI and AI models implemented for predicting chief complaints. PRISM, 2024.

Birgitta Taylor-Lillquist, Vivek Kanpa, Maya Crawford, Mehdi El Filali, Julia Oakes, Alex Jonasz, Amanda Disney, Julian Paul Keenan. Preliminary Evidence of the Role of Medial Prefrontal Cortex in Self-Enhancement: A Transcranial Magnetic Stimulation Study. Brain Sciences, 2020.