Next Generation Sequencing

NGS and Machine Learning

• The error rates of NGS reads range from ~0.1–10%. ^poplin

• The previous best variant calling algorithms were highly specialized to the Illumina platform, and are thus

• In 2018, engineers at Google demonstrated a convolutional neural network—a prominent class of neural network used in computer vision—named DeepVariant that could identify SNP and indel mutations more than 50% more accurately than the next best algorithm. ^poplin

  Bibliography

  [poplin] Poplin, Chang, Alexander, Schwartz, Colthurst, Ku, Newburger, Dijamco, Nguyen, Afshar & others, A universal SNP and small-indel variant caller using deep neural networks, Nature biotechnology, 36(10), 983-987 (2018). ↩