# CoDing Sequence Typer (CDST): a fast, simple, decentralized and interoperable solution for bacterial genomic typing and clustering

## TL;DR

CDST is a fast and decentralized method for bacterial genomic typing that matches traditional methods in accuracy while being much faster and using less storage.

## Contribution

A decentralized, hash-based framework for bacterial genomic typing that is faster and more storage-efficient than existing methods.

## Key findings

- CDST produced genomic distance matrices highly concordant with established typing methods like cgMLST and wgMLST.
- CDST achieved ~8× faster runtimes and reduced storage to ~4% of original FASTA size in a 100-genome benchmark.
- Unsupervised clustering identified three resolution levels that align well with conventional typing schemes.

## Abstract

We present a fully decentralized MD5 hash-based framework that indexes predicted coding sequences (CDSs) and computes pairwise genomic distances without locus annotation or a central database. The method is implemented in the open-source CoDing Sequence Typer (CDST) pipeline, which delivers reproducible, privacy-preserving and computationally efficient bacterial typing. Applied to 1,961 complete Salmonella enterica genomes, CDST produced distance matrices that were highly concordant with core-genome multilocus sequence typing (cgMLST) and whole-genome MLST (wgMLST), core-genome SNP, Mash and Split Kmer analysis. In a 100-genome benchmark, CDST achieved ~8× faster runtimes than cg/wgMLST workflows and reduced storage to ~4% of the original assembly FASTA size. Unsupervised clustering evaluation identified three optimal resolution levels, HC67 (outbreak-level), HC186 (lineage/serotype-level) and HC441 (global structure-level), that align well with conventional typing schemes. These levels demonstrated high internal cohesion and external consistency with conventional typing schemes. Cross-species validation on Listeria monocytogenes and Escherichia coli genomes confirmed that the pipeline recovers species-specific population structures without parameter adjustment. Collectively, CDST provides a scalable and interoperable framework for bacterial population structure analysis, suitable for surveillance and outbreak investigations across laboratories. The CDST pipeline, along with all evaluation scripts, is openly available on GitHub at https://github.com/l1-mh/CDST.