Pratapa
Cell Biology & Discovery AI
Duke University, Durham NC
Hello! I'm Aditya (Ady). Welcome to my corner of the internet!
I build computational methods that integrate multimodal data to understand how cells organize within tissues, and how that spatial organization shapes health and disease. I am an Assistant Research Professor in the Department of Cell Biology and the Discovery AI initiative at Duke University. Before that, I was a postdoctoral associate at Duke with Rohit Singh and Purushothama Rao Tata, where I developed SAME for aligning multimodal spatial data and USHER for adapting foundation models to new data.
Prior to Duke, I was a Senior Data Scientist at Akoya Biosciences, where I built computational tools for multiplexed spatial imaging of proteins and RNA. As a postdoctoral fellow at the Broad Institute of MIT and Harvard, I worked with Juan C. Caicedo on image-based cell phenotyping. I received my PhD in Computer Science from Virginia Tech, advised by T. M. Murali, where I developed BEELINE and was named the department's PhD Student of the Year in 2020. Earlier, I earned an M.S. in Computational Science at IIT Madras, developing Fast-SL with Karthik Raman and Shankar Balachandran.
Outside of research, I spend my downtime photographing birds (mostly hummingbirds) and doing some deep-sky astrophotography. More recently, I've gotten into oil painting, partly as an excuse to spend less time in front of a screen.
Multimodal integration of spatial transcriptomics, proteomics, and metabolomics.
Adapting pretrained models for out-of-distribution biological data.
Benchmarking and inference of gene regulatory networks from single-cell data.
Optimal experimental design strategies for discovery.
USHER
Aligns foundation-model embeddings across labs and protocols with fused Gromov-Wasserstein optimal transport, without retraining. RECOMB'26
Paper →SAME
Topology-flexible "space-tearing" transforms that unify protein, RNA, and metabolite data across serial tissue sections. Under revision
Preprint →BEELINE
The benchmark that exposed the limits of popular gene regulatory network methods. Cited in over 1,000 works. Nature Methods
Paper →Research
Education
