Measurement of uncertainty for STR allele sizing

General recommendations

Based on the assessments of measurement of uncertainty described below, it is recommended that all expansion-positive results, as well as normal alleles close to the threshold (+/2 repeat units) and close to the sequencing read length (e.g. 49 repeats of trinucleotide repeat unit), if deemed as possibly clinically relevant, are assessed by reviewing STR pileup plots (see STR visualisation), and, if required, an orthogonal test is performed.

For detailed recommendations for each STR loci please refer to NHS England Guidelines for Rare Disease Whole Genome Sequencing & Next Generation Sequencing Panel Interpretation & Reporting.

The accuracy and measurement of uncertainty of STR repeat allele sizing by ExpansionHunter depends on overall allele length.

Alleles shorter than sequencing read length

For alleles shorter than a sequencing read length, i.e. where the whole repeat and some flanking regions on both sides can be spanned by a single sequencing read (i.e. <150bp), the estimate of allele size by ExpansionHunter is very accurate.

In comparison with the allele sizes obtained through standard of care testing for the same samples, 98% of the allele sizes from ExpansionHunter were within +/-2 repeat units of the sizes reported by the standard of care tests, and 92% within +/-1 repeat unit.

In many cases, manual inspection of sequencing reads through visualization software (REViewer plots) revealed that ExpansionHunter was likely to estimate the allele size more accurately than standard of care tests. In a small number of cases where the detected difference from the standard of care test was >2 repeat units, ExpansionHunter typically overestimated the allele size, suggesting that it is more likely to produce false positive than false negative calls.

Alleles longer than sequencing read length

When allele length exceeds a sequencing read length, the allele sizing accuracy of ExpansionHunter is reduced.

Validation against standard of care methods showed that clinically relevant STR expansions can still be detected with high sensitivity (see Appendix G for up to date values of STR detection performance), however estimates of exact repeat length are less accurate.