Fig. 3

Collisions affect group-based variant calling methods. A Schematic showing read group-based variant calling in the presence of collision. In the absence of exogenous barcodes (top), variant information is lost due to colliding reads without the variant outnumbering those with the variant. Exogenous barcodes allow tagging of the original input molecules (bottom) and preserve this information for the sequencing reads, leading to successful detection of the variant. B Schematic representation of the experimental setup used to investigate collisions in different types of input material; 3 types of input (FF, FFPE and cfDNA) were generated from the reference sample GM12891; various input amounts were used to generate libraries that were subsequently enriched using hybridization capture with a panel targeting 110 loci and then sequenced; resulting FASTQ files were optionally merged to obtain higher number of total sequenced molecules for each input type; after sequencing read alignment, variant calling was performed using read mapping positions alone or using mapping positions and UMI information. C Collision rates plotted for three different types of input DNA (with input amount ranging from 5 to 50 ng of cfDNA, 5–200 ng of FFPE DNA and 5–300 ng of FF DNA) as a function of the number of sequenced molecules (estimated using mapping position and UMIs). D For the three types of input DNA, plots showing the number of molecules calculated using mapping positions alone (cyan) or mapping position and UMI information (red), as a function of DNA input amount. E Collision rate as a function of sequencing depth for three different types of DNA. Plots in (C), (D) and (E) were generated from FASTQ files obtained by merging all files obtained for a given type of input