Past studies on the
genetics of ASD (autism spectrum disorder) have suggested that mutations in
protein-coding genes only account for 30 percent of the spontaneous mutations with no established family history. In
the past, there have been speculations about the role of non-coding DNA in ASD;
but no studies have been conducted screening the entire genome to look for
mutations in regulatory DNA which may contribute to autism.
Troyanskaya et al., used
a novel method where they trained a machine learning model to predict how a
stretch of non-coding or junk DNA can affect gene expression in autism. They
applied this model to the Simon-Simplex collection which is an autism sample
population with the whole genomes of 2000 families with four people including
unaffected parents, sibling and affected individual. These 2000 families had no
previous history of autism so the affected individual is probably a case of
spontaneous mutation.
Jian Zhou, PhD,
co-author
of the study at the Flatiron Institute's Center for Computational
Biology (CCB); the siblings acted as a control for the experiment. The team
used the
computer algorithm on 1790 quartets. The system was able to
learn the patterns in the genome, identify relevant sections of the DNA and
also predict if mutations in non-coding DNA actually caused autism.
The final effect is the
generation of a '
disease impact score' generated by the
computational system giving an estimate of how likely a mutation in junk DNA
can have an effect on the disease.
The interesting aspect
of these computational results is the fact that mutations in the non-coding
regions affected genes and functions which had been previously linked to autism
through studies on coding genes. This finally links to a causal factor of
noncoding junk DNA mutations to ASD etiology.
The research suggests
that
junk DNA mutations affect the expression of genes in the
brain including those genes directly responsible for neurodevelopment and
neuron migration. The study is the first functional study to
examine the role of junk DNA mutations and link them up with
neurodevelopmental disorders like
autism.
Prof. Troyanskaya said that the study allows for serious analysis of 98 percent of the human genome which was not thought to contribute to protein-coding.
The study shows the importance of looking at noncoding junk DNA to understand the regulatory
mechanisms of genes and causal factors of diseases. The same method and
framework can also be used to study the role of non-coding mutations
responsible for cancer or other chronic disorders.
The paper was published
in
Nature Genetics.
References:
- Whole-genome deep-learning analysis identifies contribution of noncoding mutations to autism risk. - (http://dx.doi.org/10.1038/s41588-019-0420-0)
- Artificial intelligence detects a new class of mutations behind autism - (https://www.princeton.edu/news/2019/05/28/artificial-intelligence-detects-new-class-mutations-behind-autism)
Source: Medindia