Fig. 5: Locityper can accurately genotype mucin and other gene families.

a, Gene model of MUC1, a mucin tethered to the surface of epithelial cells. MUC1 harbors a 20-amino-acid VNTR repeat sequence and is highly polymorphic in VNTR length⁵⁹, as represented by the example haplotypes 1–3. b, Gene model of MUC5B, a secreted, gel-forming mucin that is important for homeostasis in the lungs. MUC5B encodes an irregular 29-amino-acid VNTR motif that is broken up into separate VNTR domains by cysteine domains. The number of VNTR domains, cysteine domains and VNTR motifs could each contribute to polymorphism among haplotypes at this locus⁶⁰. c, Difference in average haplotyping accuracy (QV) between Locityper and the 1KGP call set at 15 mucin genes based on 39 Illumina WGS datasets. Improvement for the LOO setting and the full Locityper database are shown using dark and light shades, respectively. Tethered and secreted mucins are shown in purple and green; the only non-gel-forming secreted mucin MUC7 is marked with an asterisk. d, Locityper (LOO) and 1KGP call set average genotyping accuracy (QV) across four gene families: CFH (orange); CYP2 (light green); FCGR (red); and MUC (blue). The diagonal black line shows the zero improvement boundary and the diagonal gray lines show a QV improvement of 10, 20 and 30. PTS, proline (P), threonine (T), serine (S).