Deep-Learning Model Can Detect Disease-Causing Mosaic Mutations

Published in Nature Biotechnology, researchers from the University of California San Diego School of Medicine and Rady Children’s Institute described the creation of a deep-learning model that could spot mosaic mutations at a faster rate than human medical geneticists.

Despite the wide range of capabilities of current DNA mutation software detectors, most cannot perceive mosaic mutations, that is, DNA mutations in a small percentage of cells, existing within normal DNA sequences, according to a press release. In addition, medical geneticists reviewing DNA sequences by eye can be time-consuming and may lead to error.

However, amid the recent spike in healthcare artificial intelligence (AI) tools, researchers found that deep learning showed high potential in learning by example from large amounts of data and performing necessary tasks. Deep learning is a type of machine learning that involves the use of artificial neural networks to process visually represented data, much like how humans visually process data, the press release noted. 

Training the deep-learning model, called DeepMosaic, involved researchers supplying it with previous examples of mosaic mutations and normal DNA sequences, training it to differentiate between them.  The researchers used increasingly complex datasets to train the model, enabling it to identify mosaic mutations better than humans and previous methods. They tested the model on numerous independent large-scale sequencing datasets.

“DeepMosaic surpassed traditional tools in detecting mosaicism from genomic and exonic sequences,” said study co-first author Xin Xu, a former undergraduate research assistant at UC San Diego School of Medicine and now a research data scientist at Novartis, in the press release. “The prominent visual features picked up by the deep learning models are very similar to what experts are focusing on when manually examining variants.”

According to senior study author Joseph Gleeson, MD, Rady Professor of Neuroscience at UC San Diego School of Medicine and director of neuroscience research at the Rady Children’s Institute for Genomic Medicine, detecting mosaic mutations is the first step to developing treatments for various disorders, including epilepsy.

“Epilepsy affects 4 percent of the population, and about one-quarter of focal seizures fail to respond to common medication. These patients often require surgical excision of the short-circuited focal part of the brain to stop seizures.  Among these patients, mosaic mutations within the brain can cause epileptic focus,” said Gleeson in a press release. “We have had many epilepsy patients where we were not able to spot the cause, but once we applied our method, called ‘DeepMosaic,’ to the genomic data, the mutation became obvious. This has allowed us to improve the sensitivity of DNA sequencing in certain forms of epilepsy, and had [sic] led to discoveries that point to new ways to treat brain disease.”

Deep learning is increasingly being applied to clinical care. In November, researchers from Massachusetts General Hospital and Brigham and Women’s Hospital said they had created a deep-learning model that could predict the 10-year risk of death from a heart attack or stroke. They used the deep-learning model to analyze X-ray images to supply clinicians with insights into the risk of cardiovascular disease.

More research from August found that a deep-learning tool could help neuroradiologists diagnose brain tumors. Developed using MRI scan data, researchers trained the tool to evaluate each image to characterize various types of intracranial tumors. They then tested the diagnostic accuracy of the model on several datasets, concluding that it achieved high accuracy, sensitivity, and specificity.