SCNA calling

We used multiple methods to assess putative cellularity and ploidy. Firstly, CNVkit was performed to generate segmented logR calls for tumour-normal pairs, and ABSOLUTE was used to estimate tumour purity, ploidy and absolute copy number per segment. The ABSOLUTE solutions were reviewed manually according to the recommended best practice. Beside, as the majority of ubiquitous mutations are likely clonal, their VAFs (after correcting for copy number) should reflect the tumour purity, says the maximum peak of variant allele fraction distribution should equal to one half of the purity. ASCAT/Battenberg estimates of purity and ploidy were also included to compare with ABSOLUTE solution. Therefore, we assessed whether the tumour purity and ploidy esimates from copy number analysis (ABSOLUTE/ASCAT/ Battenberg) were consistent with VAF estimates among multi-regions. We excluded samples with low purity and the VAF peak was <0.1, or samples had no concordant solution.

To estimate somatic copy number alternations, ASCAT/Battenberg was performed on multi-regional paired tumour-normal sequencing data. Allele counts of positions from 1000 genomes were generated using AlleleCounter, and minimum coverage of 20 for normal sample was used for filtration.  LogR and BAF values were produced for each region, and concatenated into one matrix separately for each patient. LogR values were subsequently corrected using a GC wave correction implemented in ASCAT, and only heterozygous BAF values were reserved for further analysis. Allele-specific multi-sample segmentation was performed to generate segmented logR and BAF data by ascat.asmultipcf. Both ASCAT and Battenberg algorithms were applied to provide cellularity (purity) and ploidy estimates with gamma setting of 1, and manual verification was used to select the optimal model for ploidy and cellularity using an orthogonal measures based on ABSOLUTE results and mutation variant allele fraction. And then ASCAT/Battenberg was re-run to obtain the final allele-specific copy number data using reviewed cellularity and ploidy. To assess SCNA calls, we compared absolute copy number profiles using four software packages – ABSOLUTE, CNVKit, ASCAT and Battenberg. All the samples showed high concordance except patient 12 R1 and R4, which had no best solution in ABSOLUTE.