Crowdsourcing for the assessment of somatic variant detection using synthetic tumour genomes. Homo sapiens

Two synthetic tumour-normal WGS datasets generated by VarBen were used for performance evaluation. In the first evaluation (Evaluation sample 1, ES1), a synthetic tumour-normal pair was simulated based on existing high-depth (~60×) WGS data for a blood sample from a healthy individual. A total 4409 variants (including 1693 SNVs, 1274 Indels, and 1442 SVs) were introduced with 80% tumour cellularity and three subclones. In the second evaluation (Evaluation sample 2, ES2), a high-depth (~60×) BAM file derived from the NA12878 cell line was used to generate a less complex synthetic tumour-normal pairs, containing 3556 variants (including 1324 SNVs, 1566 Indels, and 666 SVs) with 90% tumour cellularity and no subclonality. All variants introduced into the sequence data were actual somatic mutations reported in the COSMIC database.