MuSE

An accurate and ultra-fast somatic mutation calling tool for whole-genome sequencing (WGS) and whole-exome sequencing (WES) data from heterogeneous tumor samples. This tool is unique in accounting for tumor heterogeneity using a sample-specific error model that improves sensitivity and specificity in mutation calling from sequencing data.

Introduction

Detection of somatic point mutations is a key component of cancer genomics research, which has been rapidly developing since next-generation sequencing (NGS) technology revealed its potential for describing genetic alterations in cancer. We previously launched MuSE 1.0¹, a statistical approach for mutation calling based on a Markov substitution model for molecular evolution. It has been used as a major contributing caller in a consensus calling strategy by the TCGA PanCanAtlas project² and the ICGC Pan-Cancer Analysis of Whole Genomes (PCAWG) initiative³.

We have now released MuSE 2.0, which is powered by parallel computing by taking advantage of multi-core resources on a machine and an efficient way of memory allocation. MuSE 2.0 takes the same input files and outputs the same results as MuSE 1.0, but achieves a 40-50x speedup compared to MuSE 1.0. For example, MuSE 1.0 takes 2-4 hours to run on a pair tumor-normal WES data, and ∼40 hours for one pair of tumor-normal WGS data. On the other hand, MuSE 2.0 can complete running one pair of tumor-normal WES data in 5-7 minutes and one pair of tumor-normal WGS data in an hour, thus removing somatic mutation calling as a time-consuming obstacle for cancer genomic studies.

Platform

1. MuSE 1.0 supports both Linux system and MacOS.
2. MuSE 2.0 only supports Linux system with gcc=7.0 and git=2.0 or above.

Installation

git clone https://github.com/wwylab/MuSE.git
cd MuSE
./install_muse.sh

The executable file MuSE will be generated in the same directory.

Pre-processing

Before running MuSE, raw WES/WGS data need to be processed with the following software, as outlined in the following flowchart. Please refer to GDC best practice guidelines (https://docs.gdc.cancer.gov/Data/Bioinformatics_Pipelines/DNA_Seq_Variant_Calling_Pipeline/) for a detailed description of the pre-processing pipeline.

1. Biobambam (https://github.com/gt1/biobambam) to convert to FASTQ if the input is a binary sequence alignment map (BAM) file
2. cutadapt (https://cutadapt.readthedocs.io/en/stable/) for adapter and low-quality bases trimming
3. bwa (http://bio-bwa.sourceforge.net/) for alignment against the reference genome
4. samtools (http://www.htslib.org/) to sort and index BAM files
5. picard (https://broadinstitute.github.io/picard/) to mark duplicates
6. GATK 3.7 (https://github.com/broadgsa/gatk) for base quality score recalibration

Input data

Same as MuSE 1.0, MuSE 2.0 requires the following files as input:

1. Indexed reference genome FASTA file.
2. BAM formatted sequence data from the pair of tumor and normal DNA samples that have gone through the pre-processing step.
3. dbSNP variant call format (VCF) file that should be bgzip compressed, tabix indexed, and based on the same reference genome.

Running MuSE

MuSE runs in two steps. The first step, MuSE call, takes Files (1) and (2) as the input. This step carries out pre-filtering and calculating position-specific summary statistics using the Markov substitution model.

Usage:   MuSE call [options] tumor.bam matched_normal.bam
Options:
         -f FILE    faidx indexed reference genome file
         -O STR     output file name (suffix '.MuSE.txt' is
                    automatically added)
         -n INT     number of cores specified (default=1)

Example: 
MuSE call -f Reference.Genome -O Output.Prefix -n 20 Tumor.bam Matched.Normal.bam

The second step, MuSE sump, takes the output file from MuSE call and File (3) as the input. This step computes tier-based cutoffs from a sample-specific error model. We provide two options for building the model, one for WES data (option -E), and the other for WGS data (option -G).

Usage:   MuSE sump [options]
Options:
         -I FILE    single input file generated by 'MuSE call'
         -G         input generated from whole genome sequencing data
         -E         input generated from whole exome sequencing data
         -O STR     output file name (VCF format)
         -D FILE    dbSNP vcf file that should be bgzip compressed,
                    tabix indexed and based on the same reference
                    genome used in 'MuSE call'

Example:
MuSE sump -I Output.Prefix.MuSE.txt -G -O Output.Prefix.vcf -D dbsnp.vcf.gz

Output of MuSE

The final output of MuSE is a VCF file (v4.1) that lists the identified somatic variants along with tiered rankings (in the FILTER field) for these mutations. The rankings range from PASS which is the highest confidence category, followed by Tiers 1-5, with Tier 5 being the tier at the lowest confidence. The INFO field of the VCF file is always SOMATIC.

Acknowledgement

We thank Mehrzad Samadi and his team from Nvidia Corporation, including Tong Zhu, Timothy Harkins and Ankit Sethia, for their contributions towards implementing accelerating techniques in the MuSE call step in MuSE2.0.

Reference

1.  Fan, Y. et al. (2016) ‘MuSE: accounting for tumor heterogeneity using a sample-specific error model improves sensitivity and specificity in mutation calling from sequencing data’, Genome biology, 17(1), p. 178. doi: 10.1186/s13059-016-1029-6.

2.  Ellrott, K. et al. (2018) ‘Scalable Open Science Approach for Mutation Calling of Tumor Exomes Using Multiple Genomic Pipelines’, Cell Systems. Cell Press, 6(3), pp. 271-281.e7. doi: 10.1016/j.cels.2018.03.002.

3.  Campbell, P. J. et al. (2020) ‘Pan-cancer analysis of whole genomes’, Nature. Nature Publishing Group, 578(7793), pp. 82–93. doi: 10.1038/s41586-020-1969-6.