Benchmarking Code for SAVANA Manuscript

Each folder contains scripts to load the SVs into an RDS and (if applicable) compare SVs from those RDS objects to each other

tumour_replicates

Comparison between split tumour bams (in-silico replicates). VCFs have been labelled with their replicate calls using savana evaluate functionality, allowing a 100bp overlap (INFO column LABEL=SOMATIC/NOT_IN_COMPARISON).

• load_tumour_replicates.R -> saves data from all callers into rds/SVs_tumour_splits_svs.rds

srWGS_lrWGS

Comparison between srWGS (GRIDSS) and lrWGS from all algorithms

• load_srWGS_SVs.R -> saves srWGS (Illumina+GRIDSS) svs into rds/illumina_svs.rds
• load_srWGS_CNAs.R -> saves srWGS (Illumina+PURPLE) CNAs into rds/illumina_cnas.rds
• load_lrWGS_SVs_all_algorithms.R -> saves lrWGS (all callers) into rds/SVs_tumour_lrWGS_svs.rds
• compare_srWGS_lrWGS.R -> uses rds/illumina_svs.rds and rds/SVs_tumour_lrWGS_svs.rds to label both srWGS and lrWGS SVs and CNAs into:
  • rds/lrwgs_svs_validation.rds -> all lrWGS algorithms SVs, with columns for "ILLUMINA_MATCH", "ILLUMINA_MATCH_SBND", and "ILLUMINA_MATCH_CNA"
  • rds/illumina_svs_validation.rds -> contains a "caller"_MATCH validation column for every caller, set to TRUE if lrWGS SV from that caller within 500bp
  • rds/illumina_sbds_validation.rds -> same as above but for srWGS single breakends
  • rds/illumina_cna_validation.rds -> same as above but for srWGS CNAs

normal_replicates

Comparison between split normal bams (replicates) from all callers

• load_normal_replicates.R -> loads and saves labelled VCFs from all callers (INFO column LABEL=SOMATIC/NOT_IN_COMPARISON) into rds/SVs_normal_splits_svs.rds