mSINGSlw

MSI calling from sequence capture experiments

A lightweight tool to replicate mSINGS without any installation or virtual environment requirements. It is written to work as it is with minimal dependencies.

version 0.1.1

Introduction

mSINGSlw is a tool to call MSI/MSS in tumour sequencing data (specifically from small capture experiments), based on repeat length distributions. The rationale being that repeat sites in MSI samples are frequently mutated, resulting a a wider distribution of lengths. If enough regions are sampled, the proportion of mutant sites is higher in MSI samples than MSS. The original paper describing this is available here. We have written the basic algorithm into this new tool in order to minimise the requirement of the user to change permissions, create virtual environments or require administrator privileges. We make no claims of novelty, just usefulness.

Scope

mSINGSlw will provide a score, based on the proportion of mutant repeat regions. For this, you need a bam file for every sample, and a list of regions to test (see below). You will also need a list of known MSS samples as controls, or a precomputed control file, made from MSS samples. No matched normal for the tumour sample is required.

Installation and requirements

mSINGSlw requires python3, with the numpy and pysam modules installed. We leave it up to the user to install these as they see fit, either with pip, conda or other means. No installation is required, just download the mSINGSlw.py python script manually, or by:

wget https://github.com/omicsForestry/mSINGSlw/mSINGSlw.py

You can even copy/paste the code from this site into your own file. If you want to, you can make the script executable.

chmod +x mSINGSlw.py

You can move into a directory in your PATH, or make the current directory part of your PATH

export PATH=`pwd`:$PATH

However, none of this is necessary. The script should run as it is now, but you will have to specify its path when calling, and possibly that it requires python.

Usage

python mSINGSlw.py [-h] [-c CONTROLS] [-p PREBUILT] [-b BUILD] [-s SAMPLES]
                   [-r REGIONS] [-o OUTPUT] [-d DEPTH]

Lightweight implementation of MSIngs

optional arguments:
  -h, --help            show this help message and exit
  -c CONTROLS, --controls CONTROLS
                        Input list of control bam files
  -p PREBUILT, --prebuilt PREBUILT
                        Prebuilt control file, created by -b option. Either -c
                        or -p must be specified.
  -b BUILD, --build BUILD
                        Filename to save control information from files
                        provided by -c option.
  -s SAMPLES, --samples SAMPLES
                        Sample bam files to test in tab separated: file,
                        sampleID format.
  -r REGIONS, --regions REGIONS
                        Regions to test. Must be specified if -c option is
                        used.
  -o OUTPUT, --output OUTPUT
                        Output file to save if -s option is used to specify
                        samples.
  -d DEPTH, --depth DEPTH
                        Minimum depth required for processing repeat

Options

• Control samples The CONTROLS option specifies a file with a list of bam files from known MSS samples eg:

pathToControlBam1.bam
pathToControlBam2.bam
pathToControlBam3.bam
...

• Prebuilt controls If you want to use the same controls multiple times, you can save the control data with the BUILD option. If you have already saved control data with the BUILD option, you can reload it with the PREBUILT option, and leave out the CONTROLS option. The more control samples you provide, the better your calling will be.

• Test samples The SAMPLES option specifies a file with a list of bam files to test, and their sample names eg:

path2TestBam1.bam	sample1
path2TestBam2.bam	sample2
path2TestBam3.bam	sample3
...

If you are merely loading CONTROLS samples and saving with the BUILD option, then you do not need to specify SAMPLES.

• Test regions To either make a prebuilt control using the BUILD option or testing SAMPLES against CONTROLS, you will need to specify the REGIONS to test. This option is a filename with information about repeat regions. We have used the format described by MSIsensor-pro using their scan option:

chromosome	location	repeat_unit_length	repeat_unit_binary	repeat_times	left_flank_binary	right_flank_binary	repeat_unit_bases	left_flank_bases	right_flank_bases
chr1	155407852	1	0	12	943	1004	A	TGGTT	TTGTA
chr1	167411405	1	3	11	958	81	T	TGTTG	ACCAC
chr1	182177045	1	3	11	802	23	T	TAGAG	AACCT

We provide an example REGIONS file in the "examples" folder here, but you will need to make your own depending on your sequencing experiment.

• Output The OUTPUT option specifies an output text file to be made. This will provide a list of filenames, and a proportion of mutated sites. Our experience suggests that a score of above 0.2 is generally indicative of MSI, but that will depend on your regions.

• Depth The DEPTH option specifies sequence depth below which to not process a region. The default is 30, but you may wish to change this to be more or less stringent. If your data is fairly good quality, lower depth may be fine. If your data is messy, you might want to impose a higher threshold.

Example usage

We provide example MSS and MSI samples, in bam format, alongside a precomputed control file in the "examples" folder. The test bam files are listed in the samples.txt file. To run mSINGSlw on these samples:

python mSINGSlw.py -p preCompControls.txt -s samples.txt -o msi_output.txt

The following output file is made:

sample	score
sample1	58.333333333333336
sample2	0.0

This file indicates that the MSI sample1 has 54% of its repeat regions mutated, whilst the MSS sample2 has none of its regions mutated.