A Nextflow pipeline to process LCM data post-CaVEMan or Pindel variant calling
hairpin wrapper by CASM IT at the Sanger institute
[vcfilter]() - a wrapper by CASM IT (public repo coming soon)
MutationsPy - no need to install independently as it has become a git submodule
Clone this repository, including the MutationsPy submodule
git clone --recursive git@github.com:Phuong-Le/process_sanger_lcm-nf.git"""
sample_paths: a file that contains paths to relevant files for the processes in the pipeline, \
with headers similar to demo_files/lcm_processing_input_indel.tsv for Indels \
or demo_files/lcm_processing_input_snv.tsv for SNVs. \
NOTE that the data_dir column is not neccessary. \
the Simple Name of the files should be the same as the sample ID \
vcf filenames need to end with "vcf.gz" \
Due to cgpVaf requirements, the bam files should be named "${sample_id}.bam" \
the bai files should be named "${sample_id}.bam.bai" \
vcfilter_config: pre-cgpVaf filtering, default to data/indel_default.filter for Indels and data/snv_default.filter for SNVs
mut_type: either "snv" (default) or "indel"
reference_genome: absolute path to reference genome fasta file
high_depth_region: absolute path to high depth region bed file
reference_genome_cachedir: cachedir for reference genome - if this directory is empty the pipeline will split sequences from reference_genome, make sure this directory has reference file for all values 'CHROM' column in the produced VCF file
mutmat_kmer: the kmer size to construct the mutation matrix, default to 3
outdir: path to out directory
"""
nf_script=path/to/process_sanger_lcm-nf/main.nf
config_file=path/to/config/file
sample_paths=path/to/lcm_processing_input_snv.tsv
mut_type=snv
reference_genome=absolute/path/to/genome.fa
high_depth_region=absolute/path/to/HiDepth_mrg1000_no_exon_coreChrs_v3.bed.gz
reference_genome_cachedir=path/to/genome/cachedir
mutmat_kmer=3
outdir=path/to/out/directory
nextflow run ${nf_script} -c ${config_file} \
--sample_paths ${sample_paths} \
--mut_type ${mut_type} \
--reference_genome ${reference_genome} \
--high_depth_region ${high_depth_region} \
--outdir ${outdir}"module load nextflow/23.10.1-5891
module load hairpin/1.0.6
module load vcfilter/1.0.4
module load tabix/1.13
module load cgpVAFcommand/2.5.0
module load R/4.1.0
working_dir=path/to/working/directory
nf_script=path/to/process_sanger_lcm-nf/main.nf
config_file=path/to/process_sanger_lcm-nf/sanger_lsf.config
sample_paths=path/to/lcm_processing_input_snv.tsv
mut_type=snv
reference_genome=/lustre/scratch124/casm/team78pipelines/canpipe/live/ref/human/GRCh38_full_analysis_set_plus_decoy_hla/genome.fa
high_depth_region=/lustre/scratch124/casm/team78pipelines/canpipe/live/ref/human/GRCh38_full_analysis_set_plus_decoy_hla/shared/HiDepth_mrg1000_no_exon_coreChrs_v3.bed.gz
reference_genome_cachedir=path/to/genome/cachedir
mutmat_kmer=3
outdir=${working_dir}
bsub -cwd ${working_dir} -o %J.${mut_type}.out -e %J.${mut_type}.err -R "select[mem>5000] rusage[mem=5000]" -M5000 \
"nextflow run ${nf_script} -c ${config_file} --sample_paths ${sample_paths} --mut_type ${mut_type} --reference_genome ${reference_genome} --high_depth_region ${high_depth_region} --outdir ${outdir}"
Step 1: Filtering based on CaVeMan or Pindels. (default filtering criteria can be found in either data/indel_default.filter or data/snv_default.filter)
Step 2: for each donor with multiple samples, use cgpVaf to calculate VAF across their samples.
Step 3: Use beta-binomial test based on VAF to filter out germline and LCM artefact mutations.
Step 4: Generate mutation matrix (only implemented for snps at the moment, will add this feature for indels)
Phuong Le (email: al35@sanger.ac.uk)
This pipeline (particularly the sanger_lsf.config) has only be tested on non-malignant colon data (usually 2000-9000 mutations per sample). If you have a bigger dataset, it might be neccessary to experiment with different memory settings.
If neccessary, future work might \
Thanks to Chloe Pacyna and [Yichen Wang] for guiding me through the necessary steps to process LCM data
Thanks to Shriram Bhosle from CASM IT for showing me how to use cgpVaf (ie vafCorrect)
The BetaBinomial filtering of germline and artefacts was adapted from Tim Coorens's code at https://github.com/TimCoorens/Unmatched_NormSeq/tree/master
The validatation of parameter was adapted from code by Harry Hung