SquiggleNet: real-time, direct classification of nanopore signals

Yuwei Bao; Jack Wadden; John R Erb-Downward; Piyush Ranjan; Weichen Zhou; Torrin L McDonald; Ryan E Mills; Alan P Boyle; Robert P Dickson; David Blaauw; Joshua D Welch

doi:10.1186/s13059-021-02511-y

SquiggleNet: real-time, direct classification of nanopore signals

Genome Biol. 2021 Oct 27;22(1):298. doi: 10.1186/s13059-021-02511-y.

Authors

Yuwei Bao¹, Jack Wadden^{1

2}, John R Erb-Downward³, Piyush Ranjan³, Weichen Zhou⁴, Torrin L McDonald⁵, Ryan E Mills^{4

5}, Alan P Boyle^{4

5}, Robert P Dickson^{3

6

7}, David Blaauw³, Joshua D Welch^{8

9}

Affiliations

¹ Department of Computer Science and Engineering, University of Michigan, Ann Arbor, 48109, MI, USA.
² Department of Electrical and Computer Engineering, University of Michigan, Ann Arbor, 48109, MI, USA.
³ Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, 48109, MI, USA.
⁴ Department of Computational Medicine and Bioinformatics, University of Michigan, Ann Arbor, 48109, MI, USA.
⁵ Department of Human Genetics, University of Michigan Medical, Ann Arbor, 48109, MI, USA.
⁶ Department of Microbiology and Immunology, University of Michigan Medical School, Ann Arbor, 48109, MI, USA.
⁷ Michigan Center for Integrative Research in Critical Care, Ann Arbor, 48109, MI, USA.
⁸ Department of Computer Science and Engineering, University of Michigan, Ann Arbor, 48109, MI, USA. [email protected].
⁹ Department of Computational Medicine and Bioinformatics, University of Michigan, Ann Arbor, 48109, MI, USA. [email protected].

Abstract

We present SquiggleNet, the first deep-learning model that can classify nanopore reads directly from their electrical signals. SquiggleNet operates faster than DNA passes through the pore, allowing real-time classification and read ejection. Using 1 s of sequencing data, the classifier achieves significantly higher accuracy than base calling followed by sequence alignment. Our approach is also faster and requires an order of magnitude less memory than alignment-based approaches. SquiggleNet distinguished human from bacterial DNA with over 90% accuracy, generalized to unseen bacterial species in a human respiratory meta genome sample, and accurately classified sequences containing human long interspersed repeat elements.

Keywords: Deep learning; Oxford Nanopore; Raw signal; Read-until; Real-time.

Publication types

Research Support, N.I.H., Extramural

MeSH terms

DNA, Bacterial / analysis
Deep Learning*
Humans
Long Interspersed Nucleotide Elements
Metagenome
Nanopore Sequencing / methods*
Respiratory System / microbiology

Substances

DNA, Bacterial

Abstract

Publication types

MeSH terms

Substances

Grants and funding