Deep learning mutation prediction enables early stage lung cancer detection in liquid biopsy


Nov 07, 2017 (modified: Nov 07, 2017) ICLR 2018 Conference Blind Submission readers: everyone Show Bibtex
  • Abstract: Somatic mutation calling at ultra low allele frequencies is an unmet challenge that is intractable with current state-of-the-art mutation calling methods. The ability to detect cancer associated mutations in ultra low allele frequency is a fundamental requirement for cancer early detection, monitoring, and therapy nomination using liquid biopsy methods (cell-free DNA). Here we defined a spatial representation of sequencing information adapted to be used in a convolutional architecture that enables variant detection at allele frequencies of $10 \times ^{-5}$. This is 2 orders of magnitude below the current state of the art. We validated our method on both simulated plasma and on clinical plasma samples from cancer patients and healthy controls. This method introduces a new domain within bioinformatics and personalized medicine, somatic whole genome mutation calling for liquid biopsies.
  • TL;DR: Current somatic mutation methods do not work with liquid biopsies, we apply a CNN architecture to a unique representation of a read and its ailgnment, we show significant improvement over previous methods in the low frequency setting.
  • Keywords: somatic mutation, variant calling, cancer, liquid biopsy, early detection, convolution, deep learning, machine learning, lung cancer, error suppression, mutect