Researchers from the University of California San Diego School of Medicine and Rady Children’s Institute for Genomic Medicine have published a study in which they describe a method for using Deep Learning artificial intelligence to detect mosaic mutations. This involves using artificial neural networks to process data in a manner similar to human visual processing, but with higher accuracy.

The accuracy of detecting mosaic mutations is important in the development of treatments for various diseases. Researchers trained their Deep Learning model called DeepMosaic to detect mosaic mutations by providing examples of both trustworthy mosaic mutations and normal DNA sequences, teaching it to distinguish between the two.

The model was trained and retrained using increasingly complex datasets and was eventually able to identify mosaic mutations more accurately than both human and other previous methods. When tested on large-scale sequencing datasets that it had not been trained on before, the DeepMosaic still outperformed prior approaches.

A prime example of the value that DeepMosaic can bring is Focal Epilepsy. Mosaic mutations within the brain are a cause of focal epilepsy, which affects 4% of the population and can cause seizures that don’t respond to common medications, often requiring surgery to remove the affected part. With DeepMosaic, the researchers were able to improve the sensitivity of DNA sequencing in certain forms of epilepsy, leading to a variety of potential new treatment options for these kinds of brain diseases.

DeepMosaic is an open-source platform that allows other researchers to train their own neural networks to detect mutations in a targeted way using a similar image-based setup and is freely available to scientists.