A Recursive Strategy for Deep Homology Discovery in Genetic Sequence Alignment

Modern bioinformatics pipelines are engineered for speed, statistical confidence, and taxonomic consistency. These design choices enable large-scale annotation but systematically suppress the anomalies most informative for evolutionary discovery. These classification-optimized workflows treat cross-taxonomic matches and long, moderate-identity alignments as errors to be filtered rather than signals to be explored. This paper introduces a discovery-oriented methodological framework built around two core innovations: Long Alignment Cross-Taxonomic BLAST, which prioritizes structural conservation over sequence identity; and REBLAST (Recursive Evolutionary BLAST Profiling), a recursive strategy that uses anomalous alignments to guide iterative database construction and hypothesis refinement. Together, these approaches reorient sequence similarity search away from static classification and toward dynamic, evidence-driven exploration. By emphasizing long alignments, excluding close relatives, and recursively mining anomalous hits, the framework recovers deep evolutionary relationships missed by standard tools. This includes signals of ancient horizontal gene transfer, structural domain mobility, and cross-lineage conservation patterns. A companion case study demonstrates E-value = 0 alignments over 2,000 bp between human sequences and marine invertebrates, supporting the framework’s practical utility. Designed for researchers seeking to explore rather than confirm, this methodology expands the analytical territory accessible to comparative genomics, molecular evolution, and metagenomics—where the unexpected is not noise, but the path to discovery. Project data is availabe on OSF: https://osf.io/69whf/