GeneMark

GeneMark is a family of computational tools used for ab initio gene prediction in DNA sequences. Developed to identify protein-coding genes, the software has been applied to a range of genomes, from bacteria and archaea to more complex eukaryotes and metagenomic data. The programs rely on probabilistic models, primarily Markov chains and hidden Markov models, to distinguish coding regions from non-coding sequence and to model gene structure such as start and stop signals and, in eukaryotes, exon–intron boundaries.

Over time, GeneMark has expanded into several variants designed for different data types and genome complexities. GeneMark.hmm uses hidden Markov models to capture gene structure and sequence signals in prokaryotic genomes and some eukaryotes. GeneMarkS introduced self-training capabilities, enabling gene prediction without a manually curated training set, which improved applicability to newly sequenced organisms. GeneMarkS-2 further enhances self-training and performance across a wider range of eukaryotes and metagenomes. GeneMark-ET combines ab initio prediction with external hints from RNA sequencing data to improve accuracy in eukaryotic genomes. These variants are often used within annotation pipelines to provide initial gene models that can be refined by other tools.

GeneMark has had a broad impact on genome annotation workflows and is frequently cited in genome projects and comparative studies. It is distributed as standalone software and available through various public and institutional resources, sometimes alongside web-based interfaces. The name reflects its core use of Markov models to delineate coding regions and gene structure within genomic sequences.