GNN Theory & Methods
Statistical principles for graph neural networks and graph algorithms with interpretable, reproducible pipelines.
Methodology
Canonical Correlation Analysis
High‑dimensional estimators exploiting sparsity and graph structure; confidence measures for downstream decisions.
Topic Modeling for count data
Spectral methods for sparse, spatial, and tensor count data, with applications from scientific documents to microbial communities.
Uncertainty Quantification
Coverage‑calibrated UQ for structured estimators and GNNs to support reproducible scientific claims.
Structured Estimation
High‑dimensional estimators exploiting sparsity and graph structure.
Machine Learning
Principled selection and filtering of training data to improve learning while controlling unintended changes in model behavior.
Applications
Spatial Transcriptomics & RNA Velocity
Expression and velocity as complementary signals; unsupervised GNN embeddings and spatial PCA reveal spatial programs.
Photosynthetic Microbes
Multimodal integration for thermotolerance and phototaxis; linking omic profiles to stress phenotypes.
Microbial Communities
Topic models reveal microbial community structure across environments, while mechanistic studies examine how bacterial vitamin sharing emerges from release and uptake.
Brain Connectome Prediction
Understanding when graph convolutions help brain connectome prediction, and how bias, measurement error, and double-dipping affect evaluation.