A University of Toronto team has developed a machine-learning tool that can predict whether a certain genetic variant will affect RNA splicing. As it reports in Science, the group examined more than 650,000 intronic and exonic variants to search for patterns of abnormal splicing due to mutations.
The team developed a computational model using deep learning techniques that can look at input sequences and by its features or cis-elements predict what transcripts will then be expressed in a certain cell, as GenomeWeb's Uduak Grace Thomas reports.
"[We] used machine learning to essentially learn a computational model" whose output "mimics the biochemistry of the cell," senior author Brendan Frey from the University of Toronto tells her.
Such a tool, the Globe and Mail notes, could not only help identify disease-causing mutations, but also ones that make people healthier.
In Science, Frey and his team report that they were able to use their tool to find some 39 genes with splicing aberrations linked to autism, 19 of which they said were compelling ASD candidate genes.
"Computers have been used to read the genome for quite a while, but this is using a computer to interpret and understand the genome," Frey tells the Globe and Mail. "Our system is not perfect, but it works very well."