Low-bandwidth and non-compute intensive remote identification of microbes from raw sequencing reads

TL;DR

This paper introduces a low-bandwidth, non-compute intensive system for identifying microbes from raw sequencing reads by querying a remote server with minimal data transfer, enabling efficient organism identification.

Contribution

It presents a novel approach that allows microbial identification without prior reference genome specification, using a distributed architecture for minimal data transfer and computation.

Findings

System can identify microbes with minimal data transfer

Implemented web server indexing thousands of genomes

Client can run in a web browser on modest devices

Abstract

Cheap high-throughput DNA sequencing may soon become routine not only for human genomes but also for practically anything requiring the identification of living organisms from their DNA: tracking of infectious agents, control of food products, bioreactors, or environmental samples. We propose a novel general approach to the analysis of sequencing data in which the reference genome does not have to be specified. Using a distributed architecture we are able to query a remote server for hints about what the reference might be, transferring a relatively small amount of data, and the hints can be used for more computationally-demanding work. Our system consists of a server with known reference DNA indexed, and a client with raw sequencing reads. The client sends a sample of unidentified reads, and in return receives a list of matching references known to the server. Sequences for the references can be retrieved and used for exhaustive computation on the reads, such as alignment. To demonstrate this approach we have implemented a web server, indexing tens of thousands of publicly available genomes and genomic regions from various organisms and returning lists of matching hits from query sequencing reads. We have also implemented two clients, one of them running in a web browser, in order to demonstrate that gigabytes of raw sequencing reads of unknown origin could be identified without the need to transfer a very large volume of data, and on modestly powered computing devices. A web access is available at http://tapir.cbs.dtu.dk. The source code for a python command-line client, a server, and supplementary data is available at http://bit.ly/1aURxkc.