GREAC: An Open-source software for Genome Region Extraction and Classification

Background: Advances in sequencing technologies open opportunities for research into genetic variants, disease associations, and the identification of genomic signatures of natural selection that shape how organisms respond to their environments, as well as for the detection and characterisation of viral variants and the development of improved diagnostic tools. However, the analysis relied on alignment methods that compare sequences by identifying homologous regions, which poses a computational bottleneck. Thus, alignment-free techniques have the potential to maintain the quality of results whilst addressing scalability and explainability in sequence analysis. Results: This study presents GREAC (Genomic Region Extraction and Classifier), a novel k-mer-based computational methodology for identifying discriminative genomic regions that reduces dimensionality while improving classification performance. The results indicate that selected k-mers are suitable information sources for viral genome analysis, including mutation behaviour, by comparing k-mer frequency profiles and classification methodologies. The GREAC method also exhibits robust performance on classification tasks, achieving high accuracy and demonstrating strong data generalisation across varying training set sizes. Besides, the GREAC effectively reduces dimensionality by identifying discriminatory regions within biological sequences. In organisms such as SARS-CoV-2 and Monkeypox, dimensionality reduction is evident through the selection of specific coding genomic regions that retain the most informative mutations. These regions can reveal signals of speciation or population differentiation, helping to understand the evolutionary mechanisms behind divergence. Conclusions: Focusing on informative regions rather than entire sequences, GREAC addresses the critical need for explainability and transparency in genomic results. Compared to competitor methods, GREAC offers transparency and explainability in its pattern discovery by directly extracting discriminative k-mers from raw genomic data and transforming them into representative signals. This strategy enhances the interpretability of genomic features while eliminating dependence on reference genomes. GREAC’s open-source implementation ensures replicability and scalability, offering a robust tool for bioinformatics research and the development of tailored biotechnology applications, openly accessible at https://github.com/SALIPE/GREAC.