Scientific Background¶

The MUC1 Gene¶

The MUC1 gene, located on chromosome 1q22, encodes mucin-1, a transmembrane glycoprotein expressed on the apical surface of epithelial cells, including kidney tubular cells (Kirby et al., Nat Genet 2013). A defining feature of MUC1 is a large coding Variable Number Tandem Repeat (VNTR) in exon 2.

Variable Number Tandem Repeats¶

VNTRs are stretches of DNA where a short sequence motif is repeated in tandem. The number of copies varies between individuals. In MUC1, each repeat unit is approximately 60 bp (encoding 20 amino acids), and individuals carry between 20 and over 125 copies per allele (Saei et al., iScience 2023; Kirby et al., Nat Genet 2013).

ADTKD-MUC1¶

Autosomal Dominant Tubulointerstitial Kidney Disease caused by MUC1 mutations (ADTKD-MUC1) is characterized by progressive tubular damage and renal fibrosis, typically leading to end-stage kidney disease in adulthood (Bleyer et al., Kidney Int 2014). The causative mutations are frameshift insertions or deletions within the VNTR coding sequence that produce a truncated, misfolded protein termed MUC1-fs (Dvela-Levitt et al., Cell 2019). The most common mutation is a cytosine duplication (dupC) at position 67 of the repeat unit (Kirby et al., Nat Genet 2013).

The Genotyping Challenge¶

Standard short-read aligners (BWA, Bowtie2) rely on unique mapping positions. Within the highly repetitive VNTR, reads map ambiguously to multiple repeat units, resulting in poor alignment quality and unreliable variant calls. This causes pathogenic frameshift mutations to be routinely missed by conventional variant-calling pipelines (Saei et al., iScience 2023).

VNtyper 2's Approach¶

VNtyper 2 uses the Kestrel variant caller (Audano et al., Bioinformatics 2018), which reconstructs local haplotypes directly from k-mer frequency spectra instead of aligning reads to a reference. This bypasses the alignment bias inherent in repetitive regions. VNtyper 2 further refines raw Kestrel calls with empirical scoring thresholds, depth-based confidence classification, and optional cross-validation via the alignment-based adVNTR tool (Park et al., iScience 2022).