of R, Theta and GC scores for both concordant and discordant variants revealed a pattern of discordant variants located close to the borders of the variant clusters (Fig. 5). Importantly, similar clusters are used in the Illumina GenomeStudio software to assign genotypes to variants (AA, AB and BB), and, therefore, any genotype lying far from the cluster centre may be mistakenly assigned an irrelevant designation. This explanation for the observed discordance pattern was, however, relevant only to sample_002 and sample_003. In sample_001, distribution of discordant variants within the clusters was more or less even. Therefore, we conclude that the observed phenomenon requires further examination. Potential finding of the clustering pattern for discordant calls may then be exploited for quality control.
