Pangenomes aid accurate detection of large insertions and deletions from targeted sequencing: the case of cardiomyopathies.

Calling of variants of 20 bp and above from short-read data aligned to a pangenome graph (GRAF) instead of a linear reference genome

Gene panels reliably detect short coding variants but miss a large share of bigger ones. The authors analysed 1,969 cardiomyopathy cases and 1,805 controls sequenced with the Illumina TruSight Cardio panel, comparing a pangenome-based workflow (GRAF) with five conventional orthogonal methods — GATK HaplotypeCaller, GATK-gCNV, ExomeDepth, Manta and Lumpy-SV — for variants of at least 20 bp. After lab-based validation by PCR and Sanger sequencing, GRAF combined the highest precision and recall (F1 score 0.86) against 0 to 0.57 for the other methods. On the HG002 reference sample from Genome In A Bottle, GRAF also outperformed the other tools on exome data (F1 0.97 versus 0 to 0.94) and slightly exceeded GATK HaplotypeCaller on small variants of 1 to 19 bp (F1 0.975 versus 0.968). The authors conclude that pangenome-based workflows improve detection of large variants from targeted sequencing in the clinical setting.

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

The gain lands exactly where panels are blind: insertions and deletions above 20 bp, currently picked up by a separate MLPA or chromosomal microarray. Switching reference rather than stacking on yet another caller is a realistic pipeline decision, and the fact that GRAF does not degrade small variant calling is what makes it acceptable in routine. Two limits: the demonstration covers a single panel and a single phenotypic field, and an F1 of 0.86 still leaves a meaningful share of calls to adjudicate manually.

Analysis by Dr Thibaut Benquey