LumosVar2

Calls somatic SNVs, indels, and allelic copy number jointly across multiple samples from the same patient. These can be standard tumor/normal pair, longitudinal samples, primary/met, etc. Can also be used for tumor only calling with one or more tumor samples, ideally with a high tumor content and a low tumor content sample.

Citation

Joint analysis of matched tumor samples with varying tumor contents improves somatic variant calling in the absence of a germline sample. bioRxiv 364943; doi: https://doi.org/10.1101/364943

Prerequisites

System Requirements

• Linux OS (tested on Centos 7 and Ubuntu 17)

Dependencies

• bcftools and htslib (tested with 1.2-1.8) http://www.htslib.org/download/
• Matlab Runtime MCR 9.4 (or MCR 9.0 for releases 1.0-1.03) http://www.mathworks.com/products/compiler/mcr/
• pyyaml and libYAML https://pyyaml.org/wiki/PyYAMLDocumentation
• GSL - GNU Scientific Library https://www.gnu.org/software/gsl/

A note to TGen dback users, dependencies may be loaded on dback by sourcing setup.sh

Bam preperation

• Bams should be created using bwa-mem http://bio-bwa.sourceforge.net/
• Bams should be indexed by samtools

VCFs

• SNP VCFs can be downloaded from 1000 genomes, Exac, or similar population genotyping projects
• SNP VCFs must have the population allele frequency as "AF" in the INFO field
• Cosmic VCFs can be downloaded from cosmic or other cancer mutation database
• Cosmic VCFs must have "CNT" in the info field indicating count of samples having mutation
• There must be one VCF per chromsome for both SNP and COSMIC VCFs
• VCFs must be bgzipped and tabix indexed

Overview

Running lumosVar involves two main steps

1. normalMetrics: analyzes a set of unmatched controls to find average read depths and position quality metrics.
   IMPORTANT - unmatched controls must be generated using the same exome capture as the tumors
2. lumosVarMain: call somatic, germline, and copy number variants

Example dataset

http://tools.tgen.org/Files/lumosVar

Notes on pileup engine

• lumosVar uses a custom pileup engine https://github.com/tgen/gvm
• precompiled binary is provided for the pileup engine gvm so building is not required

Normal Metrics

• Inputs to normal metrics are defined in a yaml file, see template. You will need to edit:

###input files
bamList:   BAMLIST      ###path to file contining paths to bams
regionsFile: BEDFILE    ###path to bed file defining regions targeted in exome
refGenome: REFGENOME    ###path to reference genome that was used to align bams
snpVCFpath: VCFPATH     ###path to vcfs containg population frequencies (one for each chromosome),
                        ###including part of filename before chromosome number
snpVCFname: VCFNAME     ###filename/extension of population vcf following chromosome number

###user inputs
sexList: SEXLIST        ###comma deliminated list of sex for each bam in the bamlist in the same order
gvmPath: GVMPATH        ###path to folder containing gvm executable
outfile: OUTFILE        ###path and name of output

• The bamList file should contain absolute paths to the unmatched control bams with one per line.

• In order to correctly handle the sex chromosomes, the sex of the individuals in the bams list must be given as input. We have provided a helper script to determine sex from the bams if they are not known. This script takes the same yaml file as input as the normal metrics, but only the following fields in the "input files" section are needed.

python guessSex.py controlsConfig.yaml

The output will be written to <BAMLIST>.guessSex.txt. The last line of the output contains the sexList for the yaml file

• Normal metrics is run using the runNormalMetrics.py. It takes two input arguments, the yaml and the chromosome. It needs to be run separately on each chromosome.

python runNormalMetrics.py controlsConfig.yaml 21

LumosVar Main

• Inputs to lumosVar are defined in a yaml file, see template. You will need to edit:

#### Input Files
bamList: BAMLIST        ###path to file contining paths to bams
regionsFile: BED        ###path to bed file defining regions targeted in exome
snpVCFpath: VCFPATH     ###path to vcfs containg population frequencies (one for each chromosome),
                        ###including part of filename before chromosome number
snpVCFname: VCFNAME     ###filename/extension of population vcf following chromosome number
NormalBase: NMETRICS    ###path and filename of output form NormalMetrics step before chr number
cosmicVCF: COSMICVCF    ###path to cancer mutation count VCF
refGenome: REFGENOME    ###path to reference genome that was used to align bams

#### User Inputs
outName: OUTNAME        ###path and filename base for output files
outMat: OUTMAT          ###path and filename for "mat" file (matlab data) export
gvmPath: GVMPATH        ###path to folder containing gvm executable
workingDirectory: WORK  ###poth to folder containing files in the "work" directory in the github repo
NormalSample: NINDEX    ###position in bamList of normal sample, 0 indicates tumor only
priorF: PRIORF          ###vector of expected tumor fractions with one value per bam
                        ###for example [0.1;0.7] for a pair of bams with low and high expected tumor content
numCPU: CORES           ### number of parallel processors

• The bamList file should contain absolue paths to the tumor bams with one per line. All of the bams should come from the same patient.

• It is important that the length of priorF matches the number of bams in the bam list. We recommend the following values for priorF:

  • solid tumor: 0.7
  • purified tumor or cell line: 0.99
  • tumor adjacent normal tissue: 0.1
  • normal tissue unlikely to have tumor contamination: 0.01 (you may also set NormalSample to the indicate postion of your normal sample in your bam list file to run in matched normal mode)

To run lumosVar

./lumosVarMain lumosVarConfig.yaml

LumosVar Output

• .lumosVarSNV.vcf - somatic and germline SNV/indel calls
• .lumosVarSeg.vcf - copy number calls by segment
• .somaticPass.txt - table of somatic variant calls
• .lumosVarParam.txt - parameters used in the lumosVar run
• .exonData.tsv - copy number calls by regions in bed file
• .cloneSummary.tsv - summary of clonal variant groups
• .cloneSummary.pdf - graphical summary of clonal variant groups
• .groupLinePlots.pdf - line plots by clonal variant group
• .vafPlot.pdf - plots of variant allele fractions
• .cnaPlot.pdf - plot of copy number states
• <SAMPLENAME>.qualMetrics.tsv - quality metrics for candidate variant position
• .mat - matlab workspace export