Meet DeepMosaic: A Computer Program that ‘Learns’ to Identify Mosaic Mutations Causing Disease

Scientists from the University of California San Diego School of Medicine and Rady Children’s Institute for Genomic Medicine have developed a method “DeepMosaic” for identifying mosaic mutations using deep learning. The process involves training a model to analyze large amounts of genomic data and recognize patterns associated with mosaic mutations. The researchers hope that this approach will help increase our understanding of the genetic basis of disease and lead to the development of more effective treatments. 

Genetic mutations can lead to a wide range of disorders that are often difficult to treat or understand. One type of mutation, called mosaic mutations, is particularly challenging to identify because it only affects a small percentage of cells. These mutations can cause a variety of disorders but have been challenging to detect due to their rarity.

Present like a Pro! ExpertSlides – The secret weapon for your presentations!

What is a mosaic mutation?

A mosaic mutation is a genetic change that occurs during embryonic development. Mosaic mutations are not present in every cell of the body and are often only present in a subset of cells. These mutations can arise due to errors in DNA replication or environmental factors such as exposure to radiation or certain chemicals. Mosaic mutations are thought to play a role in the development of a variety of diseases and disorders, including cancer and certain neurological conditions.

Detecting DNA mutations

Current software’s for detecting DNA mutations are ineffective at identifying mosaic mutations within normal DNA sequences. As a result, medical geneticists often have to manually review DNA sequences, which is time-consuming and prone to error. Mosaic mutations, which occur after fertilization, have been linked to various diseases.

Deep learning, a type of machine learning involving artificial neural networks, allows computers to learn from large amounts of data in a similar way to how humans learn. This approach can be used to process visually represented data and has resulted in significant advances in computational abilities, including the detection of mutations. Deep learning models can process data with high accuracy and attention to detail, similar to how the human visual system works.

According to Joseph Gleeson, senior study author, and director of neuroscience research at the Rady Children’s Institute for Genomic Medicine, focal epilepsy is an example of a disorder that has yet to be fully understood.

Epilepsy affects a significant portion of the population, and a quarter of patients with focal seizures do not respond to traditional medications. In these cases, surgical removal of the affected part of the brain may be necessary to stop seizures. Mosaic mutations in the brain have been linked to the development of epileptic focus.

Previously, in many patients, the cause of epilepsy could not be identified, but the application of DeepMosaic allowed the mutation to be easily identified. Accurate detection of mosaic mutations is essential for medical research in the development of treatments for various diseases.

DeepMosaic’s Approach

The computer program DeepMosaic was trained on 180,000 simulated and experimentally assessed mosaic mutations and was tested on additional simulated and biologically tested mutations from a variety of genomic and exomic sources.

The researchers used a machine learning algorithm to teach a computer to distinguish between mosaic mutations and normal DNA sequences. They provided the computer with examples of both and trained it to identify the differences between them. By repeatedly training the computer with increasingly complex datasets and testing it against a variety of models, they were able to improve its ability to detect mosaic mutations. When they tested the program, called DeepMosaic, on independent large-scale sequencing datasets, it outperformed previous approaches.

DeepMosaic was more accurate than existing methods of identifying mosaic mutations in genomic data. When tested on noncancer whole-genome sequencing data, DeepMosaic had a sensitivity of 0.78, specificity of 0.83, and positive predictive value of 0.96. It also doubled the validation rate on noncancer whole-exome sequencing data compared to previous best-practice methods. The visual features identified by deep learning models are similar to what experts consider when manually examining genetic variations.

DeepMosaic is an open-source platform that allows scientists to use machine learning algorithms to identify mosaic mutations. It is freely available to researchers, who can use it to train their own neural networks to detect mutations using an image-based approach. This platform enables more targeted detection of mutations compared to traditional methods.

DeepMosaic is an accurate tool for identifying mosaic mutations in non-cancerous samples and can be used as an alternative or supplement to existing methods.

Conclusion

The development of a new computer program that can identify mosaic mutations is a significant advancement in the field of genetics and has the potential to improve our understanding of the genetic basis of disease. Mosaic mutations, which occur after fertilization and are not present in every cell of the body, are thought to have a role in the development of a variety of diseases and disorders. By using machine learning algorithms to explore large amounts of genomic data, the program can identify patterns associated with mosaic mutations, making it a valuable tool for medical geneticists in the search for more effective treatments for these conditions. 

Article Source: Paper Reference | Reference Article

Learn More: