Exploring the mechanism of individual differences in autism spectrum disorders using machine learning

Autism Spectrum Disorder (ASD) is a developmental disorder associated with difficulty interacting with others, repetitive behaviors, restricted interests, and other symptoms that can affect academic or professional performance. . People diagnosed with ASD may exhibit a variety of symptoms that vary in both behavioral expression and intensity.

As a result, some people with autism often need far more support than others to complete their studies, learn new skills, and live fulfilling lives. Neuroscientists have investigated the high variability of ASD for decades in hopes of helping develop more effective therapeutic strategies tailored to the unique experiences of different patients.

Researchers at Weill Cornell Medicine recently used machine learning to investigate the molecular and neural mechanisms underlying these differences among individuals diagnosed with ASD. their paper is Nature Neuroscienceidentify different subgroups of ASD associated with different functional connectivity between the brain and symptoms, which may be related to the expression of different ASD-related genes.

“This study was a follow-up study initiated and led by the first author, Dr. Amanda Book, when she was a graduate student in the lab led by Dr. Connor Liston,” said co-senior author of the paper. One Dr. Logan Grossenick told Medical Xpress. “This is directly inspired by previous work in her Liston lab that introduced new machine learning (ML) techniques to discover different subtypes and biomarkers of depression.”

Dr. Buch, lead author of the study, explains: It is important to understand and consider this biological variability in order to individualize treatment for individuals with ASD. It is difficult to identify a cure. .”

A major aim of recent research by Dr. Buch, Dr. Grosenick, Dr. Liston and colleagues is to examine differences in behavior and brain connectivity in people with ASD and to delineate the biologically distinct subtypes of this disorder. bottom. They were unsure whether these subtypes really existed, so they used machine learning tools to analyze clinical data, search for possible recurrence patterns, and analyze the results in an independent group of people with autism. I have reproduced the findings.

“If such subtypes exist, the next question is whether differences in brain connectivity are related to differences in gene expression across the brain, and what protein networks may be involved. We wanted to see if it was possible,” Dr. Grosenick said.

To conduct the study, researchers used two publicly available datasets containing information on the behavior of various individuals diagnosed with ASD, as well as functional magnetic resonance imaging (fMRI) scans and other brain-related data sets. used the data. They also used gene expression data collected by the Allen Human Brain Atlas and other previous research efforts.

“Using machine learning, we discovered a reliably reproducible dimension of brain behavior that underlies ASD, and that individuals with ASD clustered in this space into four subgroups,” Dr. Grosenick said. explained. “Interestingly, brain connectivity that differs from non-ASD controls was quite different in each subgroup. This connectivity was explained by different gene expression patterns and protein-protein interaction networks between individuals.” Importantly, these subgroups were reliably replicated in separate data, not only in the study, but also in independent text-mining analyzes of the biomedical literature.”

Overall, the findings collected by this team of researchers suggest that there are different robust subtypes of ASD, each associated with distinct molecular signaling pathways and neural connectivity patterns. The team has so far identified four subgroups of his ASD, but further research may reveal additional subgroups.

In the future, this recent study has the potential to improve our current understanding of ASD, as well as guide the development of therapies that target the various clinical subgroups it uncovers. . Meanwhile, Dr. Book, Dr. Grossenick, Dr. Liston and their colleagues plan to continue their work in this area in hopes of better understanding the characteristic patterns they uncovered.

“Technically, machine learning approaches provide a template for future studies linking genes, brain connectivity, and behavior in psychiatry,” added Dr. Grosenick. “We now plan to extend this work to the even larger and richer emerging ASD datasets (and other disparate psychiatric diagnoses), and We are actively working on improved ML methods for this kind of subtyping.”

Dr. Buch agrees, stating: