Data and scripts for "Longitudinal structural variant phylogenies in metastatic prostate cancer" Liu et al., 2026
Description
This study asks whether structural variants (SVs) capture evolutionary and resistance signals in metastatic castration-resistant prostate cancer. We analyzed paired longitudinal tumor whole-genome sequencing (pre-BAT and on-BAT) with matched normals from COMBAT-trial subjects on bipolar androgen therapy (BAT). SVCFit estimates the structural-variant cellular fraction (SVCF) across SV classes, including on hemizygous chromosomes, clusters SVs by prevalence, and infers phylogenies and clones. Source data and workflows for "Longitudinal structural variant phylogenies define tumor evolution under therapeutic selection pressure in metastatic prostate cancer" (Liu et al., 2026). SVCFit: https://github.com/KarchinLab/SVCFit. Every figure panel and statistic (manuscript-numbered) is regenerated by a named Figure_script/ script from the cached objects here. 1. Clinical Cohort (COMBAT) -- survivor: per-subject consensus SV calls (Manta + Delly + GRIDSS via SURVIVOR) and SVCFit-annotated VCFs; annotsv: SV annotations; facet: FACETS CNV segments. -- svcfit/cluster_v3: SVCF/CCF beds, DP-GMM clusters, trees, bootstraps; per-SV beds carry phasing (no_snp) and zygosity columns. -- circos, picto_outputs (PICTograph SNV clustering), downsampling_17x (downsampled replicates + SLURM pipeline), tree_pairs_S13 (paired SV/SNV phylogenies), clinical_metadata (PSA, IHC). -- Raw sequencing and germline calls are available from the corresponding author on request; not publicly shared due to consent restrictions. 2. Autosomal Benchmark (VISOR_benchmark) — single-timepoint, 75 conditions x 30 replicates. ground_truth, input_data, replicates (rep1-30 x exp1-5), depth; output/visor_replicate_data.rds is the cached object ($all_dat, $chrx, $pooled) for the Fig 3 and Supp S3-S7/S12 scripts. 3. Hemizygous chrX Benchmark (Hemizygous_chrX_benchmark) — single-copy X with chr22 as diploid control; 45 conditions x 30 replicates. scoring_rep1-30, scoring (REP 0 + block-bootstrap summaries), truth, cn_bar10k_m, seg10k, scripts, svcfit_chrx_bundle (patch + build). 4. Longitudinal Phylogeny Benchmark (Phylogeny_benchmark) — paired-timepoint VISOR simulations evaluating truncal classification, clone frequency, and topology recovery. 5. Prostate Mixture (Prostate_mixture) — admixtures of two prostate tumors; svcfit_chrx/svcfit_chrx_all.tsv (nine 3-cluster mixtures, chrX per-SV) and svcfit_45_all.tsv (four/five-cluster, autosomal); output/prostate_replicate_data_shared_sv.rds is the cached object for Fig 3C. 6. Figure_script: R/Python scripts for every figure and statistic; run inside Figure_script/ (some source the root visor_config.R for deposit-relative paths). figures/ holds rendered panels and statistics tables (Supplementary Tables S7-S15); see figures/README.md.* Environments: reproducing figures needs only R (see README.md); upstream variant-calling pipelines use the conda environments in conda_envs/ (see ENVIRONMENTS.md).
Files
Institutions
- Johns Hopkins UniversityMD, Baltimore
Categories
Funders
- Break Through Cancer
- Prostate Cancer FoundationUnited States
- Lustgarten FoundationUnited States