Autism Protein Interaction Map Reveals Common Therapeutic Targets

Knowing which genes are linked to autism is only half the battle—the real challenge is discovering what happens at the protein level inside developing brain cells.

Researchers at the University of California, San Francisco (UCSF) mapped the physical interactions of 100 proteins linked to autism risk, identifying over 1,800 physical connections between them. Their findings demonstrate that genetically distinct forms of autism share common molecular machinery, opening new avenues for broad-spectrum therapeutics.

Moving past gene lists in autism research

Few conditions carry a genetic signature quite as strong as autism, with over 250 high-confidence risk genes being linked to the condition.

However, reading DNA sequences only tells us where a risk originates, not how it alters the brain.

“For two decades, the genetics of autism has been a great success story of modern medicine. We have identified hundreds of genes linked to autism spectrum disorder, and in some cases, we have been able to demonstrate that specific mutations in those genes are causal. But knowing a gene is involved is only the beginning,” corresponding author Dr. Nevan J. Krogan, a professor in the Department of Bioengineering and Therapeutic Sciences at UCSF, told Technology Networks.

Scientists lacked maps showing how autism risk proteins physically interact inside developing brain cells, making it impossible to see how patient mutations alter cellular machinery.

“Genes encode proteins, and proteins do the work. They form complexes, they interact with each other, they regulate every aspect of how a cell functions. If you want to understand how a mutation causes disease, you have to understand what it does to the protein the gene encodes, and what that protein is doing in the cell,” explained Krogan. “Protein–protein interaction mapping gives us that window.”

Krogan and the team aimed to map the physical connections of 100 autism risk proteins in healthy and mutated states to test whether distinct genetic risk factors converge onto shared protein pathways.

“The question that drove this work was: what actually happens next? How do mutations in these genes lead to changes in the developing brain, and how do we translate that knowledge into medicines?” said Krogan.

Mapping shared protein hubs in autism

The team extracted 100 autism risk proteins and 54 patient mutations from cells, using mass spectrometry to identify their binding partners.

“Using affinity purification-mass spectrometry, we can pull a protein out of a cell and see exactly what it is interacting with. That tells you something genetics simply cannot: not just that a gene is involved, but how and where it plugs into the molecular machinery of the brain. That is where the therapeutic opportunity lives,” said Krogan.

The team used AI models to predict the 3D shapes of these pairs, differentiating between proteins that physically bind directly to one another vs those that are merely part of the same surrounding team. They then tested these interactions in frog embryos, human stem cells, and 3D brain organoids.

The experiment identified 1,881 protein interactions, 87% of which were previously unknown.

Rather than acting randomly, these connections concentrated in stem cells building early excitatory neurons.

“We found that many genetically distinct forms of autism converge on disrupting the same shared protein complexes,” said Krogan.

Tech Spotlight: Multiomics 2026

Explore how multiomics and computational approaches integrate genomic, transcriptomic, proteomic and metabolomic data, alongside spatial technologies, to generate deeper insights into disease biology, drug discovery and precision medicine.
View Webinar

Advertisement

Multiple autism risk proteins met at key molecular hubs, such as one built around the protein DCAF7. Disrupting this hub slowed stem cell growth and reduced forebrain size in animal models.

Patient mutations also rewired protein networks. A mutation in the FOXP1 protein prevented it from locking onto its normal partner, FOXP4. Cut free, FOXP4 bound to incorrect genomic regions, triggering premature neuron growth and hyperactive electrical firing in brain organoids—an effect fully reversed by deleting FOXP4.

Designing targeted protein therapies for autism

The findings established a clear model: autism risk proteins form shared teams in healthy cells, but patient mutations actively rewired those networks in disease. This moves autism research away from endless gene lists and toward targeting specific protein surfaces.

“That was genuinely surprising, and it completely changes the therapeutic calculus. If different mutations are all breaking the same molecular machinery, then you may not need a different drug for every mutation or even each gene. You may be able to develop one drug that addresses multiple genetic forms of autism at once. That is a profound shift in how we think about treating this condition,” said Krogan.

Combining physical maps with structural computer models allowed the team to see exactly where drugs could fit.

“Several complexes stood out to us, and we are actively prioritizing them for follow-up work. What we found particularly exciting is not just that we identified the complexes, but that we can now see exactly where on those complexes the mutations do their damage,” said Krogan.

“By integrating our interaction maps with AlphaFold-based structural predictions, we can pinpoint the precise protein interfaces that are disrupted. That level of resolution is what you need to design a drug that actually works. You need to know not just what to target, but where. This study gives us that map,” Krogan added.

Challenges remain, as initial mapping occurred in basic cell lines before validation in neural tissue. Krogan noted translation was a priority for his team: “making sure that what we observe in cells and organoids reflects what is happening in the brains of people with autism. That is why we have invested heavily in validating our findings in human neural progenitor cells and brain organoids derived from patient cells,” he said.

Bring Multi-Technique Analytical Data Into One Workflow

This app note illustrates processing, assignment, and reporting with examples from nuclear magnetic resonance, chromatography, mass spectrometry, and infrared spectroscopy.
View App Note / Case Study

Advertisement

Understanding how deleting FOXP4 corrected FOXP1 errors will also require more study.

“We identified more than 1,800 protein interactions, and we cannot pursue all of them simultaneously. The science of deciding which disrupted interactions are most causally relevant to the disease, and which are most tractable as drug targets, requires a combination of deep biological understanding and rigorous experimental validation,” said Krogan.

Next, the team plans to expand their physical maps and begin testing small molecules. “We are moving on multiple fronts simultaneously. Within the protein–protein interaction platform, we are extending our mapping to additional autism risk genes and patient-derived mutations, deepening our understanding of the convergent complexes we have identified, and beginning the work of designing and screening small molecules against the most promising interfaces,” said Krogan.

“We are also expanding our collaborations across UCSF and beyond to ensure our findings are validated in the most physiologically relevant systems possible, and through our participation in ARIA's Protein-Protein Interactions Hub, we now have the resources and the partners to accelerate all of this at a scale that would not have been possible before. The goal is to move from molecular blueprint to medicine, and we are more confident than ever that we have the tools and the knowledge to do it,” Krogan concluded.

Reference: Wang B, Vartak R, Hennick KM, et al. Autism mutations rewire protein interaction networks to drive neurodevelopmental pathology. Science. 2026;393(6814):eady4523. doi: 10.1126/science.ady4523

About the interviewee:

Dr. Nevan J. Krogan is a professor in the Department of Bioengineering and Therapeutic Sciences at the University of California, San Francisco (UCSF). He obtained his undergraduate degree from the University of Regina and his PhD from the University of Toronto.

Krogan’s lab at UCSF focuses on developing and applying quantitative, systematic proteomic, and genetic approaches to study complex biological and biomedical problems. At present, the Krogan group is focused on studying cancer, infectious disease, and psychiatric disorders.