Fig 1.
Overview of the variant calling pipeline (steps marked by dashed frames are only performed in case of the variant calling pipeline with GLM).
Table 1.
Overview of the parameters investigated for the variant calling pipeline with GLM.
Table 2.
Overview of the subjects sequenced on 454, Ion Torrent and Illumina NextSeq (comparison set marked with a c, re-sequencing set marked with an r).
Fig 2.
Median coverage of the genes in the intersecting target region in the case of 454 (black), Ion Torrent (red) and Illumina (green) considering the comparison data set.
Table 3.
True- and false positive SNV calls, sensitivity and PPV considering the comparison subset (n = 9), the re-sequencing subset (n = 5) and all data (454 and Illumina: n = 15, Ion Torrent: n = 10), using the standard analysis pipleine (without GLM). Only those variants are considered that are covered by at least 20 reads.
Table 4.
Normalized relative variable importance for all parameters characterizing SNVs, considering 454, Ion Torrent and Illumina NextSeq sequencing data.
Table 5.
True- and false positive SNV calls, sensitivity (sens) and PPV considering the training subset (454 and Illumina: n = 12, Ion Torrent: n = 7) and the test subset (n = 3), comparing the standard analysis pipleine (without GLM) and the optimized analysis pipleine (with GLM). Only those variants are considered that are covered by at least 20 reads.
Table 6.
True- and false positive indel calls, sensitivity and PPV considering the comparison subset (n = 9), the re-sequencing subset (n = 5) and all data (454 and Illumina: n = 15, Ion Torrent: n = 10), using the standard analysis pipleine (without GLM). Only those variants are considered that are covered by at least 20 reads.
Table 7.
Normalized relative variable importance for all parameters characterizing indels, considering 454, Ion Torrent and Illumina NextSeq sequencing data.
Table 8.
True- and false positive indel calls, sensitivity (sens) and PPV considering the training subset (454 and Illumina: n = 12, Ion Torrent: n = 7) and the test subset (n = 3), comparing the standard analysis pipleine (without GLM) and the optimized analysis pipleine (with GLM). Only those variants are considered that are covered by at least 20 reads.