Diverse autism genes converge to disrupt shared biological pathways, study finds

A new mapping of how proteins interact in the brain suggests that many different genetic mutations linked to autism spectrum disorder end up altering the same underlying biological networks. By determining how specific mutations rewire these protein connections, the research offers a potential roadmap for finding targeted treatments for neurodevelopmental conditions. The findings were published in Science.

Autism spectrum disorder is a neurodevelopmental condition characterized by varying challenges with social communication and repetitive behaviors. The genetics underlying the condition are highly complex, with hundreds of different genes implicated. For example, a study covered by PsyPost in 2016 found that thousands of genes might contribute to the condition in different ways. Despite identifying so many risk factors, scientists have struggled to understand how such a wide variety of genetic changes leads to similar developmental outcomes.

“Over the past 15 years, large-scale sequencing studies have identified hundreds of genes associated with autism,” study co-author Nevan J. Krogan, a researcher at the Quantitative Biosciences Institute at the University of California, San Francisco, told PsyPost. “That progress was initially very exciting; in 2012, only a few genes had been clearly implicated, whereas today the list includes several hundred.”

“However, identifying genes and mutations has not, by itself, produced a sufficiently detailed understanding of autism biology or led to many new therapeutic strategies,” Krogan continued. “A list of genes is somewhat like a parts list for a machine. To understand how the machine works, and what happens when one part is altered, you also need to understand the proteins those genes produce and how those proteins interact inside cells. That was the gap we wanted to address.”

To bridge this gap, scientists look to proteins, which are the microscopic machines that carry out most of a cell’s functions. Proteins rarely act alone. Instead, they bind to one another to form larger complexes, much like interlocking puzzle pieces. A 2026 study showed how a single autism-linked mutation can physically disrupt the binding between two partner proteins, impairing how brain cells communicate. Similarly, a 2026 preprint study that has not been peer-reviewed indicated that turning off different autism-associated genes in human brain cells disrupts shared biological pathways.

Mapping these physical connections can be highly useful for medical research. In fact, research covered by PsyPost in 2022 used protein interaction networks to identify existing medications that might counteract the biological processes underlying autism.

Building on this concept, a research team that included Krogan and lead author Belinda Wang sought to map the physical interactions of proteins associated with autism risk genes at a massive scale. They wanted to see if different mutations might converge on the same biological complexes, selectively rewiring how these proteins connect rather than simply destroying their function altogether.

“Through this work, we generated what we believe is the largest protein-protein interaction map assembled for a neuropsychiatric disorder,” Krogan said. “The map provides a molecular framework for understanding how autism-associated mutations may disrupt cellular systems. Importantly, it has also revealed potential therapeutic directions that would not have been visible from genetic information alone.”

“We are now following up on several of those opportunities through drug-development programs,” he noted. “The broader goal is to move beyond cataloging autism-associated genes and toward understanding the underlying biology well enough to identify new strategies for treatment.”

The team focused on 100 genes identified as high-confidence risk factors for autism. They introduced the protein versions of these genes into human embryonic kidney cells. This cell type is widely used in laboratories because it is easy to grow and allows for large-scale analysis of protein interactions. Using a technique called affinity purification and mass spectrometry, which acts like a microscopic fishing hook to pull out a target protein along with everything attached to it, they identified which other proteins bound to their targets.

The resulting map uncovered 1,881 total interactions, 87 percent of which had never been reported before. On average, each autism risk protein interacted with 11 other proteins. The researchers observed a highly interconnected network, meaning that the different autism risk proteins tended to bind to the same central hubs.

To determine which proteins touched directly rather than indirectly, the team used AlphaFold, an artificial intelligence system that predicts the three-dimensional shapes of proteins. They then tested one of the central hubs, a protein complex involving the molecules DCAF7 and DYRK1A, in living systems. They edited the genes of Xenopus, a type of frog often used to study early development, as well as human neural progenitor cells, which are stem cells that develop into brain cells.

Disrupting this shared protein complex impaired the multiplication of the progenitor cells and resulted in a smaller forebrain size in the frog models. This provides evidence that different autism risk proteins work together in the same physical machinery to regulate early brain growth.

The researchers then looked at the impact of the disease-linked mutations themselves. They introduced 54 specific genetic mutations derived from autistic patients into 30 of the risk proteins. They repeated their protein-fishing experiment to see how these mutations changed the interaction network compared to the typical, unmutated proteins.

The mutations caused 253 interactions to be altered, with some connections strengthening and others weakening. The researchers found that completely different mutations often caused the exact same changes in the protein network. For example, three different mutations in a gene called FOXP1 all caused its resulting protein to lose its connection to a partner protein called FOXP4.

To see how this specific severed connection affects brain development, the team grew human forebrain organoids. These are miniature, three-dimensional bundles of brain tissue grown in a dish from stem cells, which mimic the early stages of human brain development.

In organoids carrying the FOXP1 mutations, the loss of the FOXP1 and FOXP4 connection led to the partner protein attaching to the wrong sections of DNA. This biological misstep caused the brain cells to mature too early. The premature maturation altered the proportion of specific types of neurons in the organoids and increased their overall electrical activity.

This overlap in downstream effects was exactly what the researchers were looking for. “The most surprising finding was the degree of biological convergence,” Krogan noted. “We began with roughly 100 genes associated with autism, but the corresponding proteins were not acting independently. Many were connected within a smaller number of shared biological networks.”

“That suggests we may not need a separate treatment for every individual genetic mutation,” he added. “A smaller number of therapies targeting common pathways could potentially benefit multiple groups of patients. That possibility has significant implications for future drug development.”

The findings provide independent researchers a new lens to interpret the thousands of rare mutations identified in recent years.

Lisa Bradley, a senior research associate at The Centre for Applied Genomics at The Hospital for Sick Children Research Institute who was not involved in the study, noted that the research complements existing genomic efforts. “This large-scale study of bona fide high-risk genes specifically looks at the very real possibility that ASD-associated mutations rewire specific protein-protein interactions,” she told PsyPost. “For me, the key takeaway is that this strategy provides a way to move from many different ASD-associated genes toward a smaller number of convergent molecular mechanisms, which otherwise could have remained hidden.”

Bradley found the study’s experimental validation using human forebrain organoids particularly persuasive. “For me, the FOXP1 story was the most convincing because they took an ASD-associated mutation from disruption of a specific protein interaction with FOXP4 through to downstream effects on gene expression and neuronal development in brain organoids,” she said. “Other ASD studies have implicated altered cortical layer development, so this provided an important proof of principle that altered protein interactions can translate into meaningful cellular neurodevelopmental consequences.”

As with all research, there are a few things to keep in mind. The initial protein mapping was conducted in kidney cells rather than brain cells. While the researchers validated key interactions in brain-specific models later, the initial environment might lack some specialized proteins found only in developing neurons.

Bradley cautioned that the findings offer early-stage evidence for molecular mechanisms, “rather than demonstrating that they are causal mechanisms.” She added, “Some of these genes and mutations could also be associated with complex ASD, including co-occurring intellectual disability, so the molecular changes identified may include broader NDD [neurodevelopmental disorder] biology rather than mechanisms unique to autism etiology.”

The study focused on 100 high-confidence autism risk genes, but hundreds more have been linked to the condition, meaning the current map represents only a fraction of the total biological picture. “In this study, we examined a small number of genes and mutations in depth using stem cells, neurons and brain organoids,” Krogan said. “We have now built a pipeline that allows us to apply the same approach more systematically and at a larger scale. Our next goal is to identify which autism-associated mutations and biological pathways are most suitable for therapeutic development.”

Before new therapies can be developed, these molecular interactions must be observed in broader behavioral contexts. “While the molecular and mechanistic evidence is very compelling and is likely to be highly relevant to complex ASD etiology, I would like to see this tested in a mammalian developmental model where social and ASD-relevant behaviors can be examined alongside learning and cognitive impairments,” Bradley said. “This would allow more direct tests of causality for the FOXP1-FOXP4 interaction and how these molecular changes ultimately translate into social behavior.”

Additionally, artificial intelligence predictions of protein structures are estimates and sometimes miss the nuance of how proteins bend and fold in a living organism. Beyond refining these models, future research could expand this network and test additional drug candidates that might stabilize these weakened protein connections. The researchers believe the principles established in this study could eventually stretch to entirely different fields of medicine.

“The concept of biological convergence may extend beyond autism,” Krogan pointed out. “We are applying these tools to other neuropsychiatric conditions, including schizophrenia and obsessive-compulsive disorder, to look for shared pathways. We are also seeing overlap between some genes and proteins implicated in autism and those studied in cancer.”

“Because drugs already exist for certain cancer-related targets, there may be opportunities to investigate whether some of those compounds could be repurposed for neurological or psychiatric conditions,” he continued. “That possibility is still at an early stage, but it opens an unexpected avenue for collaboration and therapeutic research.”

Ultimately, the research team views this extensive protein map as a foundational resource for the scientific community, moving the field past simple catalogs of genetic risks.

“This is an important step between identifying a genetic change and developing a potential treatment,” Krogan concluded. “It is not the final step, and new therapies will still require years of research and testing. That said, our study provides a clearer biological framework for identifying therapeutic opportunities that were not apparent before.”

The study, “Autism mutations rewire protein interaction networks to drive neurodevelopmental pathology,” was authored by Belinda Wang, Rasika Vartak, Kelsey M. Hennick, Yefim Zaltsman, Zun Zar Chi Naing, Benjamin J. Polacco, Ali Bashir, Manon Eckhardt, Mehdi Bouhaddou, Jiewei Xu, Nawei Sun, Micaela C. Lasser, Yuan Zhou, Justin McKetney, Keelan Z. Guiley, Pawel Gniewek, Una Chan, Naufa Amirani, Owen Griffiths, Nishant Chadha, Reshmi Tognatta, Merve Cakir, Martin Gordon, Prachi Khare, Sam Drake, Vanessa Drury, David F. Burke, Silvano Gonzalez, Sahar Alkhairy, Reuben Thomas, Stephanie Lam, Montana Morris, Ethel Bader, Mélanie Dos Santos, Anastassia V. Komarova, Maxwell Bennett, Craig Ennis, Octavio Castillo, Yvonne Lim, Robert Martin, Meghan Seyler, Tierney Baum, Rebecca Krasnoff, George Wang, Sagnik Middya, Sheng Wang, Presley Pham, Juan Arbelaez, Dexter Pratt, Sofia Bali, Shivali Chag, Julia A. Kaye, Nadir Mahmood, Lee Spraggon, Thomas Rolland, Shawn Hervey-Jumper, James S. Fraser, Thomas Bourgeron, Steven Finkbeiner, Caroline Demeret, Danielle L. Swaney, Sourav Bandyopadhyay, Trey Ideker, Pedro Beltrao, Helen Rankin Willsey, Ruth Hüttenhain, Kirsten Obernier, Tomasz J. Nowakowski, Matthew W. State, A. Jeremy Willsey, and Nevan J. Krogan.