Summary : Princeton and Simons Foundation scientists have identified four biologically distinct autism subtypes using data from over 5,000 children and a new computational method. Each subtype has unique traits, developmental paths, and genetic signatures, which could transform how we understand, diagnose, and treat autism.
GANDHINAGAR: For decades, we’ve heard that autism is a “spectrum.” That word has become part of everyday language. But if you’ve ever tried to understand autism—whether through lived experience, a diagnosis, or even social media—you probably sensed that “spectrum” doesn’t quite capture just how complicated things really are.
Some people on the spectrum are nonverbal, while others are hyperverbal. Some are brilliant at pattern recognition, while others struggle with basic executive function. Some develop typically and later regress. Others show signs as toddlers.
Now, scientists at Princeton University and the Simons Foundation have taken a massive step forward in demystifying that complexity. In a groundbreaking new study, they’ve identified not just different “traits” but four biologically distinct subtypes of autism—each with its own genetic story, behavioral patterns, and developmental timeline.
This changes everything—from how we understand autism, to how it’s diagnosed, to how support and therapy could be tailored for individuals. It’s like going from watching a fuzzy black-and-white TV to seeing the same story in 4K Ultra-HD.
🧠 A Spectrum Reimagined: The Puzzle Metaphor
For years, autism research has been like trying to put together a puzzle—but no one realized we were dealing with pieces from several different boxes. Scientists were analyzing traits like language delay, sensory issues, or repetitive behavior in isolation, hoping patterns would emerge. But that “trait-by-trait” approach masked the big picture.
Natalie Sauerwald, an associate researcher at the Flatiron Institute and co-lead author of the study, explains: “We couldn’t see the full genetic patterns until we first separated individuals into subtypes.”
So, what did they do differently?
Instead of looking at individual traits, they zoomed out. They examined over 230 phenotypic features—a mix of clinical traits, behaviors, developmental milestones, and co-occurring psychiatric conditions—from over 5,000 children with autism. That data came from SPARK, the largest genetic autism research cohort in the U.S.
Then they applied a machine learning technique called generative mixture modeling. In simple terms: an algorithm that finds patterns among complex data and clusters people into groups that share meaningful similarities.
What emerged weren’t random clusters, but four robust subtypes of autism—each verified again in a second, independent group of children from a different study. Not just behaviorally different. Biologically different.
🧭 The Four Subtypes of Autism: What They Look Like and Why They Matter
Here’s where things get really interesting. Each group wasn’t just different in how autism showed up day-to-day—they also had distinct genetic “signatures” and patterns of when those genes are active during development.
🔹 1. Social and Behavioral Challenges (Approx. 37%)
- Key traits: Strong presence of classic autism traits—social interaction issues, repetitive behaviors—but relatively “on-time” achievement of early developmental milestones like walking or talking.
- Psychiatric overlap: Higher rates of ADHD, anxiety, OCD, and depression.
- Genetics: Low levels of rare, high-impact mutations, but high polygenic risk scores (PGS) for ADHD and mood disorders. That means common, inherited genes—not major random mutations—seem to drive the biology.
- Developmental timing: The genes involved in this group are mostly expressed after birth, suggesting that the biological effects become more pronounced later in childhood.
👉 Note: These individuals often fly under the radar in early childhood but start facing increasing social-emotional and behavioral challenges as they grow.
🔹 2. Mixed ASD with Developmental Delay (Approx. 19%)
- Key traits: Noticeably delayed speech, motor development, and frequent intellectual disability. But oddly, these individuals experience fewer psychiatric comorbidities.
- Psychiatric profile: Lower incidence of ADHD, anxiety, or depression.
- Genetics: High presence of rare inherited mutations (passed down from parents), as well as de novo mutations—those that appear spontaneously. The affected genes are mostly expressed in the womb.
- Developmental timing: The genes disrupted are active in fetal and neonatal stages, aligning with very early signs of developmental delay and the earliest average diagnosis age among all four groups.
👉 Note: These children tend to be diagnosed earlier, and their challenges stem from genetic changes that affect early brain development.
🔹 3. Moderate Challenges (Approx. 34%)
- Key traits: Milder overall difficulties. Individuals still show elevated scores on autism assessments but achieve milestones on time and show few psychiatric symptoms.
- Psychiatric profile: Very low occurrence of co-occurring disorders.
- Genetics: This group is enriched for rare de novo loss-of-function mutations, but these affect moderately constrained genes—genes that tolerate change better. These would have been invisible in earlier studies that didn’t divide autism into subtypes.
- Developmental timing: Gene disruptions mostly occur prenatally, but don’t cause severe effects.
👉 Note: These individuals often fall into a “gray area”—clearly on the spectrum, but often missed in diagnostic processes due to milder or subtler challenges.
🔹 4. Broadly Affected (Approx. 10%)
- Key traits: The most severe across the board—social deficits, communication struggles, strong developmental delays, and extensive co-occurring mental health issues.
- Psychiatric profile: High rates of anxiety, mood disorders, aggression, and intellectual disability.
- Genetics: Extremely high levels of de novo mutations in highly constrained genes—genes that are critical for brain development and do not tolerate change well. These include FMRP targets, which are also involved in Fragile X syndrome.
- Developmental timing: Gene activity disruptions are seen across all stages of development, from pregnancy through childhood, in multiple brain cell types.
👉 Note: This group shows deep biological disruption that affects the brain from its earliest formation and continues across life stages.
🔬 Why the Genetic Differences Are a Big Deal
One of the most groundbreaking aspects of this study is how genetic variation aligns so clearly with observable behavior and development.
Instead of just saying “autism is genetic,” the research shows:
- Which genes are involved
- Whether they were inherited or new
- When during development they start affecting the brain
- What kind of behaviors they correspond to
That’s incredibly rare in neuroscience—and incredibly useful.
For example:
- The Social/Behavioral group is driven by common genetic variation, which means it’s influenced by many genes with small effects (like ADHD or depression often are).
- The Broadly Affected group is driven by rare mutations that cause big disruptions in essential genes.
- The Mixed ASD with Delay group is more influenced by inherited rare mutations.
- The Moderate Challenges group opens up a new window into “in-between” genes that were previously ignored.
🧬 Timing Is Everything: When Autism’s Biology Kicks In
This study also found that the timing of gene expression matters just as much as which genes are affected.
- If your disrupted genes are active in the womb, you’re more likely to show delays very early (as in the Mixed ASD and Broadly Affected groups).
- If they kick in after birth, developmental milestones may be hit on time—but behavioral and emotional symptoms emerge later (as in the Social/Behavioral group).
Understanding this “when” factor could help us predict autism earlier and more accurately, based on genetic profiles.
🩺 What This Means for Diagnosis, Therapy, and Research
1. Personalized Diagnoses
No more one-size-fits-all labels. Diagnosing someone with autism could soon involve identifying their specific subtype—which would lead to more accurate developmental tracking and better-informed families.
2. Tailored Therapies
Interventions could be designed around a person’s biological profile, not just their outward behaviors. For example:
- Someone in the Broadly Affected class might need intensive early therapies and lifelong support.
- Someone in the Moderate Challenges class might benefit more from social skill training and sensory integration.
3. Smarter Science
Instead of trying to find one explanation for all of autism, researchers can now focus on one subtype at a time, making studies more efficient and findings more precise.
4. New Tools for Other Conditions
This same model—looking at traits holistically and linking them to genetic data—could transform how we study bipolar disorder, ADHD, schizophrenia, and even learning disabilities.
🛤️ What’s Next?
The researchers emphasize this isn’t the final word—it’s the starting point.
“There may be more than four subtypes,” says Aviya Litman, a Ph.D. student at Princeton and co-author of the study. “But we now have a framework—a way to meaningfully divide and study autism in ways that reflect real biological differences.”
Future directions include:
- Larger and more diverse participant groups
- Adding digital phenotypes (like data from wearables or apps)
- Tracking individuals over time to understand how their subtype affects life outcomes
- Using whole-genome sequencing to map regulatory DNA elements—not just the genes, but how they’re turned on and off
🔚 Finally
This study doesn’t claim to explain all of autism. But it breaks through a scientific bottleneck that has frustrated families, clinicians, and researchers for decades.
It offers something people with autism and their families have long asked for: recognition that not all autism is the same, and that real, biological differences matter—not just for understanding the condition but for helping people live better lives.