New computer program ‘learns’ to identify disease-causing mosaic mutations

Gene mutations cause hundreds of unresolved and untreatable disorders. Among them, DNA mutations in a small percentage of cells, called mosaic mutations, are extremely difficult to detect because they exist in a small percentage of cells.

Current DNA mutation detection software, while scanning 3 billion bases of the human genome, is not suitable for discriminating mosaic mutation hidden between normal DNA sequences. Often, medical geneticists must examine DNA sequences visually to try to identify or confirm mosaic mutations—a time-consuming and potentially error-prone endeavor.

Written in the January 2, 2023 issue of Natural BiotechnologyResearchers from the University of California San Diego School of Medicine and the Rady Children’s Genomic Medicine Institute describe a method that teaches computers how to detect mosaic mutations using an artificial intelligence method called “study carefully.”

Deep learning, sometimes called artificial neural networks, is a machine learning technique that teaches computers to do what comes naturally to humans: learn by example, especially from large amounts of information. Compared with traditional statistical models, deep learning models use artificial neural network to process data that is presented visually. Models work in a similar way to human image processing, with greater precision and attention to detail, leading to major advances in computing power, including development mutation now.

“An example of an unresolved disorder is partial seizures“, said 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 Genomic Medicine Institute.

“Epilepsy affects 4% of the population, and about a quarter of partial seizures do not respond to conventional medications. These patients often require surgical removal of the short-circuited part of the brain to remove the short-circuited part of the brain. In these patients, a mosaic mutation in the brain can cause focal epilepsy.

“We’ve had many patients with epilepsy where we couldn’t detect a cause, but when we applied our method, called ‘DeepMosaic’, to genomic data, the mutation became This has allowed us to improve the sensitivity of DNA sequences in some forms of epilepsy, and has led to discoveries that point to new ways to treat brain disease.”

Gleeson says accurate detection of mosaic mutations is the first step in medical research towards the development of treatments for many diseases.

First author and co-author Xiaoxu Yang, PhD, a postdoctoral scholar in Gleeson’s lab, says DeepMosaic has been trained on nearly 200,000 biological variants and genome-wide simulations until “finally, we were satisfied with its ability to detect variations from data it had never encountered before.”

To train the computer, the authors provided examples of reliable mosaic mutations as well as many normal DNA sequences and taught the computer to distinguish. By continuously training and retraining with ever more complex datasets and choosing between dozens of models, the computer was finally able to identify mosaic mutations much better than ever before. human eye and previous methods. DeepMosaic has also been tested on a number of independent large-scale sequencing datasets that it has never seen before, outperforming previous methods.

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

DeepMosaic is freely available to scientists. It’s not a calculator The program, which is an open-source platform, could allow other researchers to train their own neural networks to achieve more targeted mutation detection using similar image-based setting.