Robust Cancer Mutation Detection with Deep Learning Models Using Tumor-Normal Sequencing Data
Accurate detection of somatic mutations is challenging but critical to understanding how cancer forms and progresses. Such detection is also critical for targeting more effective treatments. In this seminar, Dr. Mohammad Sahraeian, senior principal bioinformatics scientist at Roche Sequencing Solutions, gives an overview of NeuSomatic — the first deep convolutional neural network approach for detecting somatic mutations.
Dr. Sahraeian will demonstrate how NeuSomatic can outperform conventional detection approaches, both in typical and challenging situations, that involve low coverage, low mutation frequency, damaged DNA, and/or ambiguous genomic regions. He will explain how this network can be applied across multiple technologies and pipelines, including whole-genome sequencing, whole-exome sequencing, AmpliSeq target-sequencing, varying tumor/normal purities. Dr. Sahraeian will also discuss the benefits of different coverages, ranging from 10x to 2000x.
Dr. Sahraeian is a senior principal bioinformatics scientist specializing in genomic data analysis at Roche Sequencing Solutions. He is the coauthor of “Deep convolutional neural networks for accurate somatic mutation detection,” which was published in Nature Communications. Using this approach, his Roche team received best performer recognition in two categories in U.S. Food and Drug Administration’s Truth Challenge V2.
- HLA Class II Across The Cancer Genome Atlas Cancer DatasetOctober 27, 2021Cloud Computing Workshop: Cancer Genomics Cloud (CGC)October 28, 2021Cloud Computing Workshop: FireCloudNovember 04, 2021NCI Childhood Cancer Data Initiative Annual Symposium 2021November 09, 2021