01 Spatial transcriptomics
Spatial transcriptomics measures gene expression while keeping track of where each cell sits in the tissue. This makes it possible to study how cells are organized and how they interact, but the data are large and noisy, so results depend heavily on both measurement quality and the choice of method.
I develop methods that model and reduce this noise so that the spatial signal can be trusted. My work includes correcting measurement artifacts, recovering true expression patterns across tissue, and identifying spatial domains in a way that stays stable across different samples and platforms.
02 Multi-omics in biology
Modern studies measure several layers of biology at once, such as the genome, transcriptome, epigenome, proteome, and metabolome. Each layer tells part of the story, and the harder question is how to combine them into one coherent picture of how a biological system works.
I build statistical methods that integrate these data types, recover the relationships between them, and handle the sparsity and high dimensionality that come with multi-omics measurements. The goal is to move from many separate readouts to a joint, interpretable model of the underlying biology.
03 AI for bioinformatics
AI models, including large language models, are now strong at reading biological sequences such as proteins and genomes. For real use in biology, though, they need to be more than accurate: they have to run efficiently and to report how confident they are.
I study how to apply these models to biological problems in a reliable way, from protein function prediction to sequence analysis, with attention to efficiency, calibration, and interpretation. The aim is to make AI a dependable tool for biological discovery rather than a black box.
Where I am headed: as a postdoc in computational biology, I want to bring these three areas together, working toward interpretable models that combine spatial and multi-omics data to describe how cells behave.
Methods and software
WEST
Weighted Ensemble for Spatial Transcriptomics. An ensemble method for robust spatial domain identification across platforms and species.
SpaDiff
Denoising of spot-swapping artifacts in sequence-based spatial transcriptomics, posed as a diffusion inverse problem.
Replace these links with the exact repository URLs when ready.
Selected collaborations
Guo-Cheng Yuan Lab (Icahn School of Medicine, Mount Sinai) · Alabady Lab (Plant Biology, UGA) · SMART Lab (Veterinary Medicine, UGA) · IBM Thomas J. Watson Research Center · Ye Shen Lab (Public Health, UGA)