GENSCAN

GENSCAN is a computational tool for predicting protein-coding genes in genomic DNA sequences, particularly in eukaryotes. First introduced in the 1990s, it was one of the early gene-finding programs to apply a generalized hidden Markov model (GHMM) to model gene structure within DNA. The method combines signals, such as splice sites and transcription-related cues, with content models that describe coding and noncoding regions, enabling it to infer exon–intron boundaries and predict complete gene structures from a given sequence.

The program analyzes a DNA sequence and attempts to identify the most probable arrangement of exons, introns, and intergenic regions. It outputs predicted gene coordinates, exon-intron organization, and, when possible, the corresponding amino acid translation. GENSCAN was designed to operate on individual genomic sequences and can be used to annotate candidate genes in newly sequenced genomes, often serving as part of larger annotation pipelines.

Impact and limitations: GENSCAN played a foundational role in early genome annotation and influenced subsequent ab initio gene-prediction methods. It generally performs best on vertebrate-like genomes for which training data are available. Limitations include dependence on species-specific training and the inherent difficulty of predicting genes with atypical structures, very small or large introns, or extensive alternative splicing. As with many ab initio predictors, its predictions are most reliable when complemented by additional evidence such as expressed sequence tags, homology, or transcript data, and should be manually curated in comprehensive annotation projects.