#27 · Genomics

Scalable algorithms for viral discovery, read alignment, and amplicon and genome analysis

The dominant thread is algorithmic efficiency on large sequencing datasets, with k-mer and index-based strategies recurring. bronko detects viral SNPs and minor variants from reads via k-mer pseudo-mapping without alignment, scaling near-linearly. COATswga uses k-mer discovery, filtering and interval tiling to design selective whole genome amplification primer sets in parallel. Columba applies optimized search schemes and bit-parallelism for lossless Illumina read alignment, with an RLC variant reducing pan-genome memory. HAlign-G provides low-memory multiple alignment of closely related genomes, including structural variant detection. AmpWrap automates 16S and 18S amplicon analysis from Illumina and Nanopore reads in one line. The distinct element is RNA virus discovery combining two-stage protein language model screening with Foldseek structure search across more than 10,000 metatranscriptomes, reporting 44x faster screening. Its learned representations differ from the k-mer and index methods elsewhere.