Large genetic studies have uncovered thousands of links between genes and human traits, from blood cell counts to disease risk. While these associations are powerful, they often leave a key question unanswered, how does a change in a gene actually lead to a change in biology or health? Turning statistical signals into clear biological explanations remains one of the biggest challenges in modern genetics.

In this work, researchers from Stanford University propose a way to connect genes to traits through the cellular programs they control. Instead of looking at genes in isolation, the team combines two complementary data sources. One comes from loss of function analyses, which estimate how damaging mutations in a gene affect a trait across many people. The other comes from Perturb-seq experiments, which use RNA sequencing to measure how turning genes on or off reshapes gene regulation inside individual cells.

Study overview and selection of model traits

By integrating these approaches, the researchers build causal graphs, structured models that show how regulatory genes influence biological programs, and how those programs ultimately affect traits. This allows them to move beyond correlation and begin inferring direction, identifying which genes act upstream as regulators and which changes directly influence traits.

As a demonstration, the team focused on three related blood traits that share some underlying biology. Using their combined framework, they reconstructed a gene regulatory hierarchy that helps explain why certain genes influence multiple blood characteristics at once. This example shows how linking genetic effect sizes with cell based perturbation data can reveal hidden structure in complex traits.

More broadly, the approach points toward a future where genetic association results are paired with functional experiments in relevant cell types. By doing so, researchers can better explain how genetic variation leads to changes in cellular behavior and, ultimately, human traits. This kind of integrative modeling brings us closer to translating genetic discoveries into biological insight and clinical relevance.