Research programs and projects

DDKG brings information from biomedical resources into a shared graph. It connects genes, molecular relationships and phenotypes so that evidence from different datasets can be queried together to investigate human disease, including birth defects and childhood cancer.

The work includes semantic data integration: deciding what an entity or relationship means, reconciling identifiers, and retaining where the evidence came from.

DULCA models gene expression across developmental age and cell type while accounting for differences between donors. It asks when and where genes, pathways and disease-relevant processes are active during development.

The initial focus is the developing heart, within a broader effort to model development across organs.

DevMap collects knowledge about developmental pathways, trajectories and markers, with links back to the evidence. It provides a way to examine which genes are associated with particular developmental processes and how those processes relate to birth defects.

The public DevMap-CHD resource brings together congenital heart defect gene sets from publications and biomedical resources. Each gene membership retains its source.

DevGraph connects DevMap’s curated developmental knowledge with DDKG and developmental data, including DULCA. Its purpose is to relate genes and mechanisms to cell types, anatomy and developmental time, so that disease findings can be examined in the context of normal development.

This makes the relationship between a developmental observation and its supporting evidence part of the graph.

We develop and evaluate methods for representing graphs and analyzing biomedical information. MOSAIC uses graph diffusion and spectral analysis to construct numerical representations of nodes for downstream analysis.

Other work compares text-embedding methods for standardizing biomedical terminology, including the different kinds of errors those methods make.

Our physics work includes analytical and computational studies of coupled-spin systems and coherence transfer in nuclear magnetic resonance (NMR).

A published example examines how dipolar coupling transfers magnetization between two spins in a solid-state system, and what that behavior implies for NMR experiments.

Biophysics work uses physical and computational models to understand molecular systems and interpret measurements of them.

Published work includes quantum-chemical calculations of nitrogen-15 chemical-shift tensors in peptides, connecting molecular structure with the quantities measured in NMR.