DNAsegmenter

DNAsegmenter is a command‑line bioinformatics tool designed for partitioning linear DNA sequences into biologically meaningful segments. The software automatically identifies breakpoints that separate distinct functional or evolutionary regions, such as coding islands, repeat motifs, or phylogenetically conserved blocks. By generating a high‑resolution segmentation map, DNAsegmenter facilitates downstream analyses like motif discovery, comparative genomics, and genome annotation refinement.

The core algorithm employs a hybrid hidden Markov model (HMM) and Bayesian change‑point detection. It models each DNA segment by a statistical distribution of k‑mer frequencies and secondary structural signals, learning transition probabilities between segment types during a training phase. The Bayesian layer adjusts the number of segments based on the marginal likelihood, avoiding over‑segmentation. Users can supply training sets of known functional domains to tailor the HMM parameters; otherwise, DNAsegmenter defaults to unsupervised learning on the input genome.

Implementation is in C++ with optional Python bindings for scripting. Input files are accepted in FASTA or GenBank format, and output is produced as BED and GFF3 files, along with a compact summary of segment statistics. The tool runs on Linux, macOS, and Windows via the Conda package manager or Docker container. Parallel processing is supported through OpenMP, allowing efficient handling of genomes up to several gigabases.

Typical use cases include refining gene model predictions in draft assemblies, flagging regions of high GC content or repetitive elements for masking, and detecting horizontal gene transfer events by segmenting atypical G+C or codon usage patterns. DNAsegmenter integrates cleanly with existing pipelines such as Prokka, MAKER, and OrthoFinder by providing standardized segment annotations. The software is released under the MIT license, encouraging broad adoption in both academic and commercial settings.

DNAsegmenter was first described in a 2024 publication in Bioinformatics Advances and has since been cited in over fifty peer‑reviewed studies related to genome assembly quality assessment, functional annotation, and microbial phylogenetics. The source code and documentation are maintained at the GitHub repository dna-segmenter/dnasegmenter, where users can file issues or contribute new segmentation models.