Coriolis-Lite

A lightweight C++20 metagenomic classifier: the FM-Index-backed sibling of Coriolis. Built on sdsl-lite for compressed succinct indexing. Targets mobile and edge devices where RAM and storage are tight. NSF-supported under award CNS-1910193.

What it does

Coriolis-Lite classifies sequencing reads against a reference index of microbial genomes. Input is FASTA / FASTQ; output is a per-read taxonomic assignment in a Centrifuge-compatible format, so it can drop into existing pipelines unchanged. The FM-Index gives O(m) substring queries on a compressed representation of the reference, so the classifier fits in dramatically less memory than suffix-array or hash-based alternatives.

Why it exists

Most metagenomic classifiers assume a server-class machine with tens to hundreds of GB of RAM. Coriolis-Lite targets the opposite end: phones, NVIDIA Jetson-class boards, MinION-attached devices, where sending data back to a server isn’t an option. Memory efficiency is the primary constraint; FM-Index over sdsl-lite is the right primitive because it keeps the index in compressed succinct form while preserving fast substring queries.

Stack

C++20, sdsl-lite, libdivsufsort, Boost.Iostreams, LZ4, spdlog, DNAsbt. CMake build. Linux, x86_64 + ARM (Jetson via JetPack 5.x).

Co-authors

Purushotham Sirasapalli (primary), Vatsal Labh, Chris Chen, Paridhi Desai, Troy Richardson.

Status

Shipped and in active use within the SCORE Lab metagenomics pipeline as the FM-Index-based companion to Coriolis.