GMCK Cohort Analysis Pipeline
Snakemake pipeline for somatic variant calling and copy number analysis in paired primary/relapse tumour samples. Each patient contributes three sample types: germline (lymph node), primary tumour, and relapse tumour, sequenced with a targeted panel using paired-end reads with inline UMIs.
Pipeline overview
FASTQ (per sample)
└─ UMI extraction (fgbio FastqToBam)
└─ Adapter trimming (Cutadapt)
└─ Alignment (BWA mem → GRCh37/hg19)
└─ Merge aligned + unaligned BAM (GATK MergeBamAlignment)
└─ UMI-aware duplicate marking (GATK UmiAwareMarkDuplicatesWithMateCigar)
└─ Base quality score recalibration (GATK BQSR)
└─ Somatic variant calling (GATK Mutect2, tumour vs germline)
└─ Annotation (VEP v108 + CADD pre-scored)
└─ Filtering (depth ≥ 50, VAF > 5%, consequence, CADD ≥ 0.20)
└─ Copy number analysis (snp-pileup + FACETS)
Cohort (all patients combined)
└─ Tumour mutational burden (SNVs / 2.4 Mb panel)
└─ Mutual exclusivity (DISCOVER)
└─ Ploidy-adjusted CNV burden + recurrence
└─ Paired Wilcoxon signed-rank test (primary vs relapse, BH-corrected)
└─ Plotly visualisations
Software versions
Bioinformatics tools
| Tool |
Version |
Notes |
| BWA |
0.7.17 |
Alignment to GRCh37/hg19 |
| GATK |
4.2.6.1 |
UMI dedup, BQSR, Mutect2, FilterMutectCalls |
| Cutadapt |
4.1 |
Adapter and quality trimming |
| fgbio |
— |
UMI extraction (FastqToBam, MergeShards); verify version with fgbio --version |
| samtools |
1.6+ |
BAM sorting and indexing |
| VEP |
108 |
Variant annotation, GRCh37 cache |
| snp-pileup |
— |
Allele counts for FACETS; verify version |
| FACETS |
0.6.2 |
Copy number analysis (R package) |
Python packages
| Package |
Version used in study |
| Python |
3.8.12 |
| pandas |
1.5.3 |
| numpy |
1.26.4 |
| scipy |
1.14.0 |
| statsmodels |
0.14.2 |
| plotly |
5.12 |
| pysam |
— |
| pyranges |
— |
| discover |
0.9.5 |
R packages
| Package |
Version |
| R |
3.6.3+ |
| facets |
0.6.2 |
| optparse |
— |
| data.table |
— |
Installation
Python environment
conda env create -f envs/python.yaml
conda activate gmck-pipeline
Bioinformatics tools
conda env create -f envs/tools.yaml
conda activate gmck-tools
Or load via your HPC module system (e.g. module load bwa/0.7.17 gatk/4.2.6.1).
R packages
install.packages("optparse")
install.packages("data.table")
# FACETS from GitHub:
devtools::install_github("mskcc/facets", ref = "v0.6.2")
Usage
- Copy
sample_sheet_template.tsv to sample_sheet.tsv and fill in paths.
- Edit
config/config.yaml — fill in all /path/to/... entries.
- Dry-run to verify the DAG:
cd workflow
snakemake -n --cores 1
- Full run:
snakemake --cores 16
Variant filtering criteria
- PASS calls from FilterMutectCalls only
- Read depth ≥ 50 in the tumour sample
- VAF > 5% in either primary or relapse (union)
- Consequences excluded: intronic, downstream gene, upstream gene, 3′/5′ UTR, non-coding transcript exon
- CADD PHRED ≥ 0.20
- Reference genome: GRCh37/hg19
CNV class definitions (FACETS)
| Class |
Total copy number |
| Deep loss |
TCN = 0 |
| Loss |
TCN < 2 |
| CNLOH |
TCN = 2, LCN = 0 |
| Neutral |
TCN = 2, LCN = 1 |
| Gain |
TCN 3–5 |
| Amplification |
TCN ≥ 6 |
FACETS run with --max-depth 20000 to accommodate deep panel sequencing.