Step 17: Cancer Predisposition Screening with CPSR¶

What This Does¶

Screens germline variants against curated cancer predisposition gene panels to identify clinically actionable cancer risk variants. CPSR uses its own panels sourced from Genomics England PanelApp and other curated databases — these are cancer-focused and distinct from the ACMG SF list of secondary findings (which also includes cardiac and metabolic genes not covered by CPSR).

Why¶

ClinVar screening (step 6) finds known pathogenic variants, but CPSR applies ACMG/AMP classification criteria to novel or rare variants in cancer predisposition genes — catching variants ClinVar hasn't yet classified.

Tool¶

CPSR (Cancer Predisposition Sequencing Reporter), bundled inside the PCGR image

Docker Image¶

PCGR_IMAGE

Pinned in versions.env; Image versions lists the current tag.

The image is PCGR 2.3.2 from ghcr.io/sigven/pcgr (Docker Hub's sigven/pcgr stops at 2.2.5). cpsr is on its PATH. It needs a separate ref data bundle (PCGR_DATA_BUNDLE, ~7 GB) and the VEP cache of PCGR_VEP_CACHE_RELEASE; the three move together.

Prerequisites¶

1. VEP Cache¶

PCGR 2.3.2 bundles VEP 115, which requires the release-115 cache (~24 GB). This is different from the release-116 cache used by step 13. Both coexist in the same vep_cache/ directory under different subdirectories (116_GRCh38/ and 115_GRCh38/).

mkdir -p ${GENOME_DIR}/vep_cache
wget -c -P ${GENOME_DIR}/vep_cache https://ftp.ensembl.org/pub/release-115/variation/indexed_vep_cache/homo_sapiens_vep_115_GRCh38.tar.gz
tar xzf ${GENOME_DIR}/vep_cache/homo_sapiens_vep_115_GRCh38.tar.gz -C ${GENOME_DIR}/vep_cache

2. PCGR Ref Data Bundle¶

PCGR 2.x uses a ref data bundle (~7 GB for 20260620) separate from VEP:

mkdir -p ${GENOME_DIR}/pcgr_data
cd ${GENOME_DIR}/pcgr_data
wget -c https://insilico.hpc.uio.no/pcgr/pcgr_ref_data.20260620.grch38.tgz
tar xzf pcgr_ref_data.20260620.grch38.tgz
mkdir -p 20260620 && mv data/ 20260620/

This creates a 20260620/data/ directory with ClinVar, CancerMine, UniProt, and other databases. VEP cache is now mounted separately.

Upgrading from PCGR 2.2.5¶

PCGR 2.2.5 read bundle 20250314 and the release-113 cache. 2.3.2 needs both new ones: about 31 GB to download (the 7 GB bundle and the 24 GB cache). Once step 17 runs on 2.3.2, pcgr_data/20250314/ and vep_cache/homo_sapiens/113_GRCh38/ are no longer read and can be deleted.

Command¶

source versions.env   # from the repository root
docker run --rm --user root \
  --cpus 4 --memory 8g \
  -v ${GENOME_DIR}/vep_cache:/mnt/.vep \
  -v ${GENOME_DIR}/pcgr_data/${PCGR_DATA_BUNDLE}:/mnt/bundle \
  -v ${GENOME_DIR}/${SAMPLE}/vcf:/mnt/inputs \
  -v ${GENOME_DIR}/${SAMPLE}/cpsr:/mnt/outputs \
  "${PCGR_IMAGE}" \
  cpsr \
    --input_vcf /mnt/inputs/${SAMPLE}.vcf.gz \
    --vep_dir /mnt/.vep \
    --refdata_dir /mnt/bundle \
    --output_dir /mnt/outputs \
    --genome_assembly grch38 \
    --sample_id ${SAMPLE} \
    --panel_id 0 \
    --secondary_findings \
    --force_overwrite

Panel Options¶

--panel_id takes one or more of these ids, comma-separated. The list is the one PCGR 2.3.2 ships (pcgr/pcgr_vars.py); GEP means Genomics England PanelApp. 2.3 dropped 2.2.5's panel 38 (Rhabdoid tumour predisposition), so the ids from 38 up moved down by one.

Panel ID	Description
0	CPSR exploratory cancer predisposition panel (PanelApp genes, TCGA's germline study, Cancer Gene Census, other sources), the widest panel and the one this step uses
1	Adult solid tumours cancer susceptibility (GEP)
2	Adult solid tumours for rare disease (GEP)
3	Bladder cancer pertinent cancer susceptibility (GEP)
4	Brain cancer pertinent cancer susceptibility (GEP)
5	Breast cancer pertinent cancer susceptibility (GEP)
6	Childhood solid tumours cancer susceptibility (GEP)
7	Colorectal cancer pertinent cancer susceptibility (GEP)
8	Endometrial cancer pertinent cancer susceptibility (GEP)
9	Familial Tumours Syndromes of the central & peripheral Nervous system (GEP)
10	Familial breast cancer (GEP)
11	Familial melanoma (GEP)
12	Familial prostate cancer (GEP)
13	Familial rhabdomyosarcoma (GEP)
14	GI tract tumours (GEP)
15	Genodermatoses with malignancies (GEP)
16	Haematological malignancies cancer susceptibility (GEP)
17	Haematological malignancies for rare disease (GEP)
18	Head and neck cancer pertinent cancer susceptibility (GEP)
19	Inherited MMR deficiency (Lynch syndrome) (GEP)
20	Inherited non-medullary thyroid cancer (GEP)
21	Inherited ovarian cancer (without breast cancer) (GEP)
22	Inherited pancreatic cancer (GEP)
23	Inherited polyposis and early onset colorectal cancer (GEP)
24	Inherited predisposition to acute myeloid leukaemia (AML) (GEP)
25	Inherited susceptibility to acute lymphoblastoid leukaemia (ALL) (GEP)
26	Inherited predisposition to GIST (GEP)
27	Inherited renal cancer (GEP)
28	Inherited phaeochromocytoma and paraganglioma (GEP)
29	Melanoma pertinent cancer susceptibility (GEP)
30	Multiple endocrine tumours (GEP)
31	Multiple monogenic benign skin tumours (GEP)
32	Neuroendocrine cancer pertinent cancer susceptibility (GEP)
33	Neurofibromatosis Type 1 (GEP)
34	Ovarian cancer pertinent cancer susceptibility (GEP)
35	Parathyroid Cancer (GEP)
36	Prostate cancer pertinent cancer susceptibility (GEP)
37	Renal cancer pertinent cancer susceptibility (GEP)
38	Sarcoma cancer susceptibility (GEP)
39	Sarcoma susceptibility (GEP)
40	Thyroid cancer pertinent cancer susceptibility (GEP)
41	Tumour predisposition - childhood onset (GEP)
42	Upper gastrointestinal cancer pertinent cancer susceptibility (GEP)
43	DNA repair genes pertinent cancer susceptibility (GEP)

Output¶

${SAMPLE}.cpsr.grch38.html — Interactive HTML report with classified variants
${SAMPLE}.cpsr.grch38.classification.tsv.gz — Tab-separated variant classifications (gzipped; read it with zcat)
Every variant in the panel genes gets one of the five ACMG/AMP classes: Pathogenic, Likely pathogenic, VUS, Likely benign, Benign
In the TSV, CLASSIFICATION is the final class and ASSERTION_AUTHORITY says whose it is: ClinVar's, unless ClinVar has no record or a conflicted one, then CPSR's own (CPSR_CLASSIFICATION, which every variant has). PCGR 2.2.5 named the final column FINAL_CLASSIFICATION; the reports (bin/collect_summary.py) read either. The file names are the same in both versions.

Runtime¶

~30-60 minutes per genome (depends on variant count).

Notes¶

The ref data bundle (~7 GB) and VEP cache only need to be downloaded once — shared across all samples.
PCGR 2.x breaking changes: The CLI changed completely from 1.x. The old --pcgr_dir flag (which internally appended /data) is replaced by --refdata_dir and --vep_dir as separate mount points. The single monolithic data bundle is split into a smaller ref data bundle + the standard Ensembl VEP cache. Docker volume mounts changed from a single :/genome to four separate mounts for VEP, bundle, inputs, and outputs.
Data bundle freshness: The 20260620 bundle dates from June 2026 (ClinVar 2026-06, GENCODE 49). Check the PCGR releases page periodically for updated bundles — newer bundles include more recent ClinVar classifications and gene-disease annotations.
VEP cache version: PCGR 2.3.2 requires the VEP release-115 cache, while step 13 uses release-116. Both coexist in vep_cache/homo_sapiens/ (subdirectories 116_GRCh38/ and 115_GRCh38/). You need both if running both steps.
Use --panel_id 0 for the comprehensive cancer superpanel (500+ genes) — cancer-focused.
CPSR 2.3 classifies every variant in the target genes by itself; 2.2's --classify_all is gone and 2.3 refuses it. --clinvar_trust_level (0 to 4, default 0) decides when CPSR's class replaces ClinVar's; the step keeps the default.
--secondary_findings reports pathogenic / likely-pathogenic variants in the ACMG SF incidental-findings gene list (cardiac and metabolic genes beyond CPSR's cancer panels; the list itself comes with the PCGR data bundle) in a dedicated section of the report. scripts/17-cpsr.sh passes it unless you set CPSR_SECONDARY_FINDINGS=false, for when you do not want findings outside cancer predisposition: CPSR_SECONDARY_FINDINGS=false ./scripts/17-cpsr.sh your_sample. The Nextflow module always passes it.
CPSR is complementary to ClinVar screening — ClinVar finds known variants, CPSR classifies novel ones.
The same ref data bundle is used by PCGR for somatic analysis (not relevant for germline WGS).