Computational biologists at Carnegie Mellon University are working to bring order to one of modern genetics’ biggest challenges: connecting disease-linked DNA changes to the specific cells and genes they actually affect. By merging many types of genomic and cellular data into a single computational framework, the team hopes to make it easier for researchers to pinpoint drug targets for rare and complex disorders—and to do so faster and with greater confidence.
The motivation is scale and complexity. The human body is made up of hundreds of cell types, and human genomes contain tens of thousands of variants. Over the last 20 years, advances in sequencing have enabled scientists to associate roughly 50,000 to 70,000 genetic variants with traits and diseases, including conditions such as Alzheimer’s disease and characteristics like height. But these statistical links rarely reveal the biological mechanism needed for therapy development: which cell type is impacted, which gene is truly responsible, and how that altered gene activity changes cell behavior.
Martin Zhang, an assistant professor in CMU’s Ray and Stephanie Lane Computational Biology Department within the Kenneth C. Griffin School of Computer Science, is taking on that gap with a multiyear, $2.3 million award from the National Institutes of Health. The funding comes through the NIH’s Supporting Talented Early Career Researchers in Genomics program and backs an effort to identify causal targets more precisely.
A key obstacle is that disease-associated variants are typically mapped to broad genomic neighborhoods rather than single genes. These regions—often described as loci spanning about one megabase of DNA—can contain dozens of genes. Zhang said that a single locus may include 30 to 50 candidate genes, forcing researchers to make a high-stakes choice about which one to pursue. If scientists select the wrong gene, experimental follow-up and drug development may fail to affect the disease at all, consuming years of work and substantial public resources.
Zhang’s approach goes well beyond comparing DNA sequences or measuring messenger RNA, the readout that indicates which genes are active at a given time. His team aims to combine additional layers of biology: how DNA is folded and arranged inside cells, how nearby genomic regions influence one another, and what happens when particular genes are switched on or shut off in specific cell types. The idea is that when multiple, independent datasets converge on the same gene and cell type, researchers gain stronger evidence that they have found the right mechanistic target.
According to Zhang, the grant supports about 10 interconnected subprojects organized around three priorities. The first is determining which cell types are most relevant to particular disease-associated variants. The second is linking variants to the genes they regulate directly, while avoiding analytical bias. The third is building an integrated framework that can combine all of these disparate data streams.
The long-term goal is a unified computational system that can ingest multiple biological datasets simultaneously, translating them into a shared digital “language.” Instead of treating each dataset as an isolated clue, researchers would be able to evaluate a consolidated view of gene regulation, cell context, and variant effects. Zhang describes the broader ambition as turning genetic associations into actionable therapeutic insights by understanding how variants change biology—and identifying where interventions might alter disease trajectories.
The same integrative perspective may also sharpen understanding of human evolution. In a Nature Genetics study involving Zhang, researchers reported a phenomenon they call “linkage masking.” While genetic variants are often treated as acting independently under natural selection, the team found evidence that some mutations arise and persist as pairs. For traits such as height—where extremes in either direction can be disadvantageous—an increase-causing mutation can be offset by a decrease-causing mutation, allowing both to remain in the population because their combined effect cancels out.
Why it matters
Most genetic discoveries still stop short of explaining mechanism, slowing the translation from variant to treatment—especially for rare and complex diseases. A platform that unifies diverse genomic and cellular datasets could help researchers choose correct drug targets earlier, reducing wasted effort and speeding therapeutic development. The next test will be whether these integrated methods consistently identify causal genes and cell types across diseases and datasets.