Two UC San Diego teams redraw the map of what human cells can do

Two separate laboratories at the University of California San Diego published results on successive days this month that, taken together, sketch a quieter shift in how American biomedical research is being organised: less reliance on single-gene heroics, more reliance on systematic maps and repurposed biology.

On 11 July 2026 a team of engineers at the Jacobs School of Engineering reported that they had rebuilt fragments of human immune-cell membranes into tiny particles capable of fighting drug-resistant fungal infections. Two days later, on 13 July, a separate bioengineering group at the same university released what it calls the first genome-scale reference map of how individual genes shape the behaviour of human stem cells, built from a dataset of roughly 2.5 million cells. The two papers are unrelated scientifically. They share an institution, a methodology of scale, and a bet that the next decade of medicine will be built on infrastructure rather than on individual discoveries.

From single genes to a reference atlas

The stem-cell paper, published by the laboratory of Sheng Zhong and colleagues at the UC San Diego Institute of Engineering in Medicine, is in substance a protest against the way cell biology has been practised for three decades. Researchers have traditionally inferred what a gene does by switching it off in a handful of cells, watching the result, and publishing the story. The new atlas does something closer to bookkeeping: it scores the activity of every expressed gene across millions of individual cells, then uses that score to predict the function of genes that have never been individually studied.

According to the team's summary, the underlying dataset covers approximately 2.5 million human stem cells and is being released openly so that other laboratories can query it without running their own experiments. The implied claim is consequential. If a researcher anywhere can ask a database "what does this unstudied gene do in a pluripotent cell?" and receive a credible answer, the marginal cost of generating new biological knowledge falls sharply. The bottleneck shifts from wet-lab work to computational literacy.

That shift does not come free. Genome-scale references depend on standardisation: how cells are grown, how reads are aligned, how "function" is defined. The atlas inherits the conventions its authors chose, and downstream users inherit them too. The risk is not fraud but lock-in, the same quiet consolidation that has followed other infrastructure-style resources in genomics, from reference genomes to protein-structure databases.

Immune membranes as a drug

The fungal-infection paper is the smaller, sharper story. A group at the Jacobs School, working with collaborators at the Terasaki Institute for Biomedical Innovation and other partners, took the outer membranes of neutrophils, the immune cells that arrive first at a site of infection, and wrapped them around biodegradable cores to create what the team calls cell membrane-coated nanoparticles. In mouse models of fungal keratitis, an infection of the cornea, the particles reduced the fungal load more effectively than standard antifungal drugs.

The mechanism is conceptually simple: the particles present the same surface markers a neutrophil would, and fungal pathogens that have evolved to evade human immune cells find themselves, briefly, fooled by a decoy. The therapeutic promise is significant because the fungal infections in question, particularly those caused by resistant strains of Candida and Fusarium, are increasingly hard to treat with existing drugs. The World Health Organization has flagged invasive fungal disease as a growing and under-recognised threat, and the antibiotic pipeline has produced almost no new classes of antifungal compounds in forty years.

The honest caveats are also significant. Mouse corneas are not human eyes, and a particle that performs well in a controlled infection does not necessarily perform well in the messier environment of a hospitalised patient with a compromised immune system. The team has not yet reported human-trial data, and the path from a rodent result to an approved therapy typically runs a decade.

What scale buys you

The two papers, read side by side, say something about where American biomedical research is putting its money. Both efforts are labour-intensive in the conventional sense: growing cells, profiling them, manufacturing particles, running animal experiments. What distinguishes them is the volume. A 2.5-million-cell atlas is not a finding; it is a substrate for findings. A library of membrane-coated particles is not a single therapy; it is a platform from which several therapies might be derived.

This is the same logic that produced the Human Genome Project, the Allen Brain Atlas, and the protein-structure databases of the past two decades: build the resource, release it openly, and let a thousand laboratories compete to ask interesting questions of it. The model has a mixed record. It is excellent at producing data and uneven at producing treatments; the translation gap between a published atlas and an approved drug remains stubbornly wide.

It is also a model that has come under quiet strain. Public funding for the National Institutes of Health has been flat in real terms for much of the 2020s, and large reference projects increasingly rely on philanthropy, institutional cost-sharing, and industry partnerships. The UC San Diego stem-cell atlas, for instance, drew support from the National Institutes of Health and the National Science Foundation alongside university funds, a combination that is harder to assemble each year.

Stakes and what to watch

The next obvious test for the stem-cell atlas is uptake. If, by mid-2027, laboratories unconnected to the original team have used the resource to identify functions for previously uncharacterised genes and to propose drug targets, the project will have justified its cost. If the atlas becomes a citation rather than a tool, it will join a long list of well-meaning databases that the field cites but rarely queries.

For the membrane-coated nanoparticles, the next milestone is a formal preclinical toxicology package and, eventually, an investigational new drug filing with the US Food and Drug Administration. The team has not announced a timeline. Investors watching the antifungal space, which has drawn renewed interest because of rising resistance, will be looking for a partnership with a larger pharmaceutical company as a signal that the platform can be industrialised.

Both stories are small in the scheme of global biomedical research, and both are likely to be overtaken by larger trials and louder announcements within months. But they illustrate a structural choice the field is making: to invest in the maps and platforms from which the next decade of medicine will be drawn, and to accept that the maps themselves, rather than any single discovery, will be the deliverable.

This publication framed both studies as infrastructure investments rather than as discrete breakthroughs, on the view that the long-run value of systematic reference resources is more durable than the news cycle that announces them.