Fig. 2

Overview of GeDi algorithm. pGSA and aGSA denote GeDi’s primary and auxiliary suffix arrays respectively. Data derived from tumour and control NGS datasets are given red and blue colouration respectively. a First, preprocessing filters out tumour reads that exactly match the reference and removes ‘N’ characters from the input tumour and control data (red T and blue C files respectively). b Second, suffix array-based SNV detection uses a dual suffix array design to detect SNVs, including those at low allele frequency. Variant blocks are constructed. c Third, consensus pairs (labelled T,C-pairs in diagram) are constructed from variant blocks and from control reads covering the same genomic location. False positives are removed by consensus pair filtering. d Finally, SNVs are called using control consensus pairs as proxies to compute SNV genome coordinates