An artificial intelligence (AI) system has been used to detect a collective of genetic causes for autism spectrum disorder.
Researchers at Princeton University Lewis-Sigler Institute for Interactive Genomics analysed the genomes of 1,790 families with simplex autism spectrum disorder, in which one child presents with the condition but no other family members do. The AI then sorted through 120,000 genomic mutations detected in each of the participants to isolate those which were affecting the behaviour-related genes of people with autism.
The results have been published in the journal Nature Genetics.
The deep learning AI performed successive layers of analysis to pick out patterns that would be indiscernible to human analysts. It was able to teach itself how to identify biologically relevant sections of DNA and predict whether or not they played any role in the 2,000+ protein interactions known to impact gene regulation.
The system also predicts whether disrupting a single pair of DNA units would have a substantial effect on those protein interactions.
“The algorithm ‘slides along the genome’ analysing every single chemical pair in the context of the 1,000 chemical pairs around it, until it has scanned all mutations,” Princeton University researcher Olga Troyanskaya said.
This means the system is able to predict the effect of mutating each and every chemical unit in the entire genome. In the end, it reveals a prioritized list of DNA sequences that are likely to regulate genes and mutations that are likely to interfere with that regulation.
Less than 30% of the participants affected by autism had a previously identified genetic cause for their diagnosis prior to the study, but the newly found mutations are likely to significantly increase that fraction.
The AI’s predictive ability proved key, the researchers said. Previous studies into the genetic route of autism struggled to detect any significant difference in the number of mutations in the behaviour-regulating genes of people with autism versus neurotypical people. However, the Princeton researchers’ AI looked into mutations predicted to have a high functional impact and found a significantly higher number of these in affected people.
The researchers said this information could be hugely important to the family, friends and physicians of people with autism as it will help discourage them from making overly general assumptions how one person’s autism might be classified with others.
Princeton University researcher Chandra Teesfeld said: “They say that when you meet one person with autism you have met one person with autism because no cases are alike. Genetically, it seems to be the same way.”
The results of the study prove that mutations in regulatory DNA can cause a complex disease, and the researchers now believe their algorithm can be generalised to discover the genetic contributions to a wide range of conditions. The approach could be particularly helpful for neurological disorders, cancer, heart disease and many other conditions that have eluded efforts to identify genetic causes.