In the lab, 22 million immune cells map gene circuits

A gene is easier to understand when someone presses its switch. Scientists from Gladstone Institutes, UC San Francisco and Stanford University did that nearly 12,800 times in human T cells from blood donors, screening 33.4 million cells and using 22 million high-quality cells to build a map of immune gene circuits. The study was published in Cell.

The method, called Perturb-seq, combines a targeted genetic change with single-cell analysis: researchers switch off one gene at a time, then observe what happens inside each cell. Instead of relying on long-used experimental cell lines, the team worked with primary human immune cells, which retain their ability to respond to signals that activate the immune system.

That choice exposed a central fact about biology: a gene does not produce one fixed outcome in every situation. The Gladstone team found that gene circuits operate differently depending on whether T cells are resting or responding to an infection. Alex Marson, director of the Gladstone-UCSF Institute of Genomic Immunology, described the work as a step from reading the genetic blueprint to testing how targeted changes alter a cell’s state.

Concretely, the map gives researchers a way to connect genetic variants with the pathways that shape immune traits and disease risk. It could inform the design of cancer immunotherapies and the study of autoimmune conditions, while letting scientists compare cellular responses across people.

The dataset also joins Biohub’s Billion Cells Project, an effort to assemble an open-source collection of 1 billion single cells for AI models that predict cell behavior. The researchers say the context-rich data are necessary for reliable virtual-biology models: an AI trained on only one cell state may struggle with real-world responses. For now, the achievement is a laboratory map and an open research resource, not a treatment in clinical use.