Illumina trains AI on primate data to estimate risk of rare variants in humans

Millions of people have now had their genome and exome sequenced, frequently using Illumina-developed equipment, but it is still unclear how most of the genetic variations found in the research will affect them. Models may be able to work at rates and scales greater than those of humans thanks to AI. 

PrimateAI-3D was developed by Illumina using a neural network trained on sequencing data from 233 different primate species. Primate and human proteins are quite similar, therefore the sequencing results enabled the researchers to build an artificial intelligence that can anticipate human behavior by overcoming the paucity of labelled data needed to train sophisticated machine learning models. PrimateAI-3D differentiated between benign and pathogenic variations and evaluated the pathogenicity of uncommon coding variants in two publications that were published in Science.