Fragmented assemblies can manufacture spurious "accessory" gene families: a gene broken across a contig boundary is predicted as two partial proteins, neither of which clusters with the intact family, so each becomes a fake strain-private family. This report checks whether that confound is present on this run.
| Comparison | rho (raw) | rho (partial, controlling total_length) | n strains |
|---|---|---|---|
| accessory_present vs n_contigs | 0.5595 | 0.1633 | 344 |
| accessory_present vs n50 | 0.6461 | -0.0797 | 344 |
| core_missing vs n_contigs | -0.0326 | 0.1038 | 344 |
| core_missing vs n50 | -0.2018 | -0.0716 | 344 |
WARNING: |rho| exceeds 0.3 for accessory-family content against assembly quality. Between-strain accessory-genome comparisons (accessory island counts, gain/loss calls, Leiden trans-modules) are confounded on this dataset: "strain X has more accessory genes" may mean "strain X has a worse assembly", not a real biological difference. This is a diagnostic only -- no presence call or strain has been altered or excluded.
Proteins in strain-private (singleton) families sit within 1000 bp of a contig end 16.40% of the time (n=13904), against 0.60% of all 2428684 proteins genome-wide. See assembly_quality_vs_content.tsv for the per-strain breakdown.