Repeat expansions

An expansion can be unstable and grow across generations (anticipation). It often does not change a protein directly but disturbs transcription, splicing or creates toxic RNA. That is why the exact repeat count, not just its presence, is clinically decisive. Thresholds are also locus-specific, with premutation or grey zones between normal and pathogenic, so a count with an uncertainty range is more informative than a yes-no result.

A pathogenic repeat is often longer than a single sequencing read, so no read spans the repeat from both sides. Repetitive motifs also align ambiguously, which is why a standard aligner and an ordinary variant caller underestimate or miss the true length. A reliable estimate therefore needs a targeted, locus-aware method rather than genome-wide SNV logic. Genome consequently handles these loci separately from the rest of variant analysis.

Genome uses ExpansionHunter (Illumina). For each locus from a curated catalog (including HTT, FMR1, C9ORF72, FXN, DMPK, CNBP, AR and several ATXN loci) the tool builds a sequence graph from the flanks and the repeated motif and realigns the local reads against it. It evaluates three read classes: spanning reads that fully contain the repeat, flanking reads that reach into it from one side, and in-repeat reads that lie entirely within the motif and flag especially long expansions. From these it estimates a repeat count per allele and reports a genotype confidence interval. Genome reads out these genotypes together with the confidence interval and read support and presents them strictly as repeat evidence, not as small-variant evidence.