Power of AI to help tackle genetic diseases

Artificial intelligence (AI)-powered protein models combined with genome sequencing technology could help scientists better diagnose and treat genetic diseases, according to new research from The Australian National University (ANU).  

The research findings, published in Nature Communications, could improve the future of personalised medicine by harnessing the power of new data tools. 

Using AlphaFold’s AI-powered protein structure predictions, a multidisciplinary team from the ANU John Curtin School of Medical Research and the ANU School of Computing analysed genetic variations at an unprecedented scale. 

The ANU scientists, led by Associate Professor Dan Andrews, looked at every possible mutation in the entire set of proteins found in the human body, uncovering a hidden pattern that explains why some proteins are more prone to destabilising mutations than others. 

“Our study reveals that evolution has built resilience into the most essential proteins, shielding them from harmful mutations that disrupt protein stability.  Less critical proteins seem to have not evolved this inherent ability to absorb damage,” Associate Professor Andrews said. 

According to Professor Andrews, the findings reveal why the less critical genes, rather than more essential ones, often have a greater importance for genetic diseases observed among patients.  

“Genetic mutations are like the rain that all genes must endure – they are constant and unavoidable. However, not all genes, and the proteins they encode, are equally well waterproofed,” he said. 

“Some genes are so essential that they are very rarely observed with mutations in people, while others are a little less critical but are still important enough that human diseases occur when they contain mutations.”  

The research helps prioritise treatments by identifying specific genetic pathways affected by mutations. 

“It’s important to identify which genetic system is dysfunctional in a given person, which helps us potentially choose the most effective treatment,” Associate Professor Andrews said. 

“Our study applies to complex diseases with multiple mutations as it involves scoring genetic variation for its functional effects, which is crucial for identifying potentially broken genes.

“There is also potential for clinical translation and the development of AI tools to help improve patient outcomes.

“Our future goals include developing automated systems to flag effective treatment for individuals, based on their genetic and pathology data.” 

The full study has been published in Nature Communications.