Genomic data has structure at multiple scales: bases repeat in local patterns, quality scores correlate with sequence context, identifiers follow rigid naming conventions, and reference genomes provide strong priors for read alignment. A monolithic compressor treats this heterogeneous data as a single stream and must rediscover the structure that the format already declares.
Hecate decomposes genomic compression into a conditional coding problem over coupled streams. Each component — sequences, quality scores, identifiers, auxiliary fields — gets a separate codec optimized for its specific statistics, with the codecs conditioned on each other where cross-stream correlations exist. The framework is modular: codecs can be swapped independently without redesigning the architecture.
The performance profile is telling. At equivalent compression ratios, Hecate runs 2–10 times faster than established competitors. At equivalent time budgets, it compresses 5–10% better. The advantage is largest on large genomes and referential compression — exactly the cases where structural heterogeneity is most pronounced and monolithic approaches waste the most effort.
The structural point: factoring a problem into its natural components and solving each component with a specialized tool is not always better than a general-purpose approach — it depends on whether the components are genuinely separable. Genomic data is separable: sequences are statistically independent of identifier naming conventions, and quality scores have their own temporal structure. The modularity wins because it matches the data's actual factorization. When the problem decomposes, the solution should too.