Autism is defined according to a variety of behavioral symptoms, but it is precisely this variation – along with a complex genetic background – which makes it difficult to connect behavior to the underlying genes.
A new algorithm can make this challenge a little easier to solve. The algorithm, which uses a form of artificial intelligence to learn as it goes, analyzed behavioral data and has learned to recognize six genetic disorders associated with autism, according to research published February 11 in the journal of Molecular Autism.
Researchers hope to use these behavioral signatures to narrow their search for the genetic bases of ‘idiopathic autism’, for which there is no known cause.
“There was a way to assume that genetic risk factors were likely to lead to a set of distinct behavioral phenotypes, but many researchers have never formally proven what they previously proposed,” says researcher Patrick Bolton, professor of child and adolescent psychiatry at Kings College London. “That was the motivation for this project.”
Previous studies have sought ways in which different syndromes may come from shared molecular or neurological pathways, but showed no clear consensus on how.
The new robot learning technique called support vector, performed screening through large volumes of data to find recognizable patterns that can be used to subdivide a group of people. The system searched through health information of individuals diagnosed with one of six genetic disorders linked to autism: 22q11.2 deletion syndrome (also called DiGeorge syndrome), Down syndrome, Prader-Willi, Tuberous Sclerosis, Klinefelter Syndrome and chromosome 15 isodicentric or idic, which is caused by a duplication of a segment of chromosome 15.
The method was able to identify specific behavioral signatures for each syndrome. To this end an algorithm was built that can find the same types of signatures in the behavioral data of individuals with idiopathic autism.
The researchers tested their algorithm on three samples with a total of 1,261 individuals with autism and found instances of all six signatures.
“What I find most impressive about this study is that it is expanding its classification for idiopathic autism and is showing these signatures,” says Dennis Wall, associate professor of Pediatrics at Stanford University, who has used the automated learning to develop diagnostic tools for autism, but was not involved in this research.
“It suggests that we might be able to use this, or methods like this, to find a firm genotype-phenotype correlation that can act as a gravitational force to make sense of what is now a very mixed and complex picture,” says Dennis Wall.