STAT: a fast, scalable, MinHash-based k-mer tool to assess Sequence Read Archive next-generation sequence submissions

Michael Kimelman,Oleg Shutov,Christopher O'Sullivan,J Rodney Brister,Richard Lapoint,Kenneth S Katz

doi:10.1186/s13059-021-02490-0

Abstract

Sequence Read Archive submissions to the National Center for Biotechnology Information often lack useful metadata, which limits the utility of these submissions. We describe the Sequence Taxonomic Analysis Tool (STAT), a scalable k-mer-based tool for fast assessment of taxonomic diversity intrinsic to submissions, independent of metadata. We show that our MinHash-based k-mer tool is accurate and scalable, offering reliable criteria for efficient selection of data for further analysis by the scientific community, at once validating submissions while also augmenting sample metadata with reliable, searchable, taxonomic terms.

Highlights

Established in 2007, the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) accepts raw sequencing data directly from highthroughput sequencing platforms [1]
Our design starts from the MinHash principle that a random selection of the lowest valued constituent blocks in a pool after hashing represents a signature of the parent object
The k-mer database / k-mer count annotation file pair is designated “dbss,” the database sorted by taxonomy id (TaxId), with each TaxId set sorted by k-mer

Summary

Introduction

Established in 2007, the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) accepts raw sequencing data directly from highthroughput sequencing platforms [1]. The growth of data submission is exponential (doubling approximately every 12 months [2]), rendering use of computationally expensive methods, such as de novo assembly followed by alignment, impractical due to costs and limits of scale, given the time constraint of submission processing. We considered that questions about the quality of a given NGS run could reasonably be inferred from the taxonomic distribution of reads within that set, whether based on a single organism or of metagenomic design. This is often enough information to answer basic experimental or clinical questions, as well as inform decisions about the merit of subsequent resource-intensive assessment methods. Binning reads into taxonomic buckets can identify contaminating reads and reads outside of the stated

Methods

Results

Conclusion