cisVar is a pipeline to estimate pre-ChIP frequencies from pooled post-ChIP frequencies via a regression on the genotypes of the individuals in the pool. To use this code you must have a mapped BAM file for each pool (a single pool is fine) as well as genotype info for all individuals in the pool in VCF format. Additional information can be provided also, see below for information.
For an overview of this method, or to cite this work, please see the following:
Tehranchi, Ashley K., Marsha Myrthil, Trevor Martin, Brian L. Hie, David Golan, and Hunter B. Fraser. “Pooled ChIP-Seq Links Variation in Transcription Factor Binding to Complex Disease Risk.” Cell 165, no. 3 (April 21, 2016): 730–41.
This code was written by Ashley Tehranchi with minor modifications by Mike Dacre. It was produced at Stanford University, but is released here under the MIT license.
Current version: 2.0.0b3
Contents
cisVar.py
is the pipeline written in python3 and uses regression_qtls.R
(regression_qtls.R
is called inside of cisVar.py
so you will not run it
directly).
In addition, the code at scripts/combine.py
can be used to create combined
and merged (per-SNP) pandas dataframes from the outputs of cisVar.py
for
multiple samples.
A complete example Snakemake pipeline is provided in pipeline
, you should be
able to run it by modifying only the cisvar_config.json
file, see the
Snakemake section below.
Right now the default minor allele frequency filter is 0.1>MAF<0.99
. To
change these and other regression constants, edit the regressions_qtls.R
script.
Install via PyPI:
pip install cisVar
Or install from github:
pip install https://github.com/TheFraserLab/cisVar/tarball/master
Alternatively, you can just clone the repo and use it directly from there.
git clone https://github.com/TheFraserLab/cisVar.git
It should work with python 2 or 3, but python 3 is highly recommended.
Requires samtools v1.9
and bedtools v2.26.0
as well as the
following python modules (installed automatically by pip):
pandas
, numpy
, psutil
, snakemake
, wget
Additionally requires that R
with Rscript
are installed, with
the ggplot2
module installed.
To run this code on your own data you need:
- VCF(s) with genotypes and ref/alt alleles for every SNP you wish to test
- A mapped BAM file, ideally this will make use of Hornet to remove reference bias.
In addition, the following data can be helpful:
- A list of individuals to filter genotypes with (one newline separated file of individuals per pool)
- A file with ref and alt alleles for your SNPs of interest, to modify those in the genotype VCF files (BED/VCF/txt)
- A file to limit the SNPs to consider to a subset of those in the VCF (BED/VCF/txt)
Example pipeline:
cisVar.py prep -F SampleName -i individuals.txt.gz --chrom-format chr /path/to/geno/*vcf.gz
cisVar.py mpileup -F SampleName -f fastaFile -B sortedBam
cisVar.py post -F SampleName -r readDepth -a allelesFile
cisVar.py geno -F SampleName -r readDepth -i individualsFile -g genotypesFile
cisVar.py qtls -F SampleName -r readDepth -n numberIndividuals
cisVar.py tidy -F SampleName -r readDepth
scripts/combine.py {sample}.readDepth.regression.pd
A readDepth of 20 is generally optimal, the sample name can be whatever you
want, but should be unique per sample. {sample}
is a placeholder to allow
any number of samples to be combined in the last step.
The above pipeline can be automated with Snakemake.
To use, install cisVar, navigate to the root of your project, and run
cisVar.py get_snake
to copy the Snakefile and config file over . Then edit
the cisvar_config.json
file to match your needs.
You will also need to edit the Snakefile
to set the script_prep
string
to match what is needed by your system.
The following are the config options for that file:
Option | Description |
---|---|
name | A general name for this run, file prefix will be
<name>.<sample>.<read_depth> |
sample_name | The name of the sample, default is population. Used only in the combination of multiple samples. |
samples | A list of samples, can just be a list, or a dictionary of {sample:group}, the 'group' in this case allows the use of the same genotype files for multiple samples, can also be a path to a separate json file |
read_depth | An integer depth to require for each SNP to be considered |
max_cores | Used only when parsing VCFs, if you have multiple VCF files ( e.g. per chromosome), they will be parsed in parallel up to this many cores (or max avaialable on machine) |
sort_vcfs | Either 1 or 0, if 1 assumes that VCF files contain a
chr# string in the file name, and sorts the order
of files to be chr1->22,X,Y,MT. Don't use if your
VCFs don't have chr# in the name |
chrom_form at | 'chr', 'num', 'ignore': Force format of chromosome
name to be chr# or # . This ensures that all
input files have the same format. Use ignore to do
nothing. |
bams | A path to the mapped BAM files, must contain the
{sample} string (unless you only have one bam),
e.g. /path/to/{sample}.sorted.bam , {sample}
must be in samples |
cisVar | Path to the cisVar repository |
vcfs | Can be a single path (for one vcf), a list of vcfs,
or a glob string (e.g.
/path/to/vcfs/*.vcf.comm.gz ) |
genome_fa | Path to a FastA file of the genome you mapped to, single file only. |
inds | Optional: used to filter VCFs so that the genotype
files contain only the individuals in the sample,
e.g. /path/to/inds/{sample}.ind.txt.gz . Newline
separated file of individuals. |
locs | Optional: a BED/VCF/text file of SNP locations to consider, used to limit the total to be a subset of the genotype file. |
alleles | Optional: a BED/VCF/text file of alternate ref/alt alleles. Must be a subset of the genotype VCFs. If there is an entry in this file, it's ref/alt alleles will be used instead of those in the genotype file |
Note the last three files are optional, also if samples
is a dict, then the
value will be used in place of the sample. For example, if you have two samples
for the same population that are yri1
and yri2
, but they both use the
same genotype file yri.geno.vcf
, you can make samples {'yri1': 'yri',
'yri2': 'yri'}
and then yri
will be used to pick the ind, loc, and allele
files
To run on a cluster, run cisVar.py get_snake
with -x
and edit the
cluster.json
file to match your cluster environment, then run e.g.:
snakemake -j 100 --cluster-config cluster.json \
--cluster "sbatch -n {threads} -t {params.time} --mem={resources.mem_mb} -p {cluster.queue} -o {cluster.out} -e {cluster.err}" \
all
or
snakemake -j 100 --cluster-config cluster.json \
--cluster "qsub -l nodes=1:ppn={threads} -l walltime={params.time} -l mem={resources.mem_mb}MB -o {cluster.out} -e {cluster.err}" \
all
To set the maximum allowed memory per job, add the argument --resources
mem_mb=32000
. Note, this is for the whole pipeline, not per job, because
snakemake is stupid.
To also combine files, replace all
with combine
at the end of the
command.
Below are help options available on the command line for cisVar, all these steps are run by the above snakemake pipeline.
usage: cisVar.py [-h] {prep,mpileup,post,geno,qtls,tidy,get_snake} ... cisVar: Find cis QTLs based on an experimental selection method Ashley Tehranchi <[email protected]> Stanford University Version: 2.0.0b1 Created: 2015-12-12 Updated: 2018-05-16 Example usage: cisVar prep -F test_new -i individuals.txt.gz --chrom-format chr cisVar mpileup -F <SampleName> -f <fastaFile> -B <sortedBam> cisVar post -F <SampleName> -r <readDepth> cisVar geno -F <SampleName> -r <readDepth> -i <individualsFile> cisVar qtls -F <SampleName> -r <readDepth> -n <numberIndividuals> cisVar tidy -F <SampleName> -r <readDepth> -p out.dataframe.pandas -t out.dataframe.txt Note: The qtls regression step will use approximately 32GB of memory on an averaged- sized dataset. The geno step will use approximately 20GB of memory on the same dataset. positional arguments: {mpileup,post,geno,qtls} prep Prepare genotype files mpileup (m) Run mpileup post (p) Run POST frequency calculation geno (g) Munge genotypes to prepare for regression qtls (q, regression, r) Run the regression tidy (t) Tidy up regression, call open/clsoed optional arguments: -h, --help show this help message and exit
This step converts VCFs into genotype and individual files that can be used by the pipeline.
usage: cisVar.py prep [-h] [-F PREFIX_NAME] [-r TRIAL_DEPTHS] [-i ALL_INDS] [-l LIMIT_FILE] [-a ALLELE_FILE] [--chrom-format {chr,num,ignore}] [--include-indels] [-c CORES] vcf_files [vcf_files ...] Prepare genotype files optional arguments: -h, --help show this help message and exit Run Options: -F PREFIX_NAME, --SampleName PREFIX_NAME sample/population name (default: cis_var) -r TRIAL_DEPTHS, --readDepth TRIAL_DEPTHS minimum read depth per variant (default: 20) Prep Options: -i ALL_INDS, --all-inds ALL_INDS File of individuals in all groups, one per line -l LIMIT_FILE, --limit-file LIMIT_FILE BED/VCF/txt file of SNPs to consider -a ALLELE_FILE, --allele-file ALLELE_FILE BED/VCF/txt file of alleles to override VCF allels (subset of vcf) --chrom-format {chr,num,ignore} chr: make format "chr#", num: make format "#", ignore: do nothing (default: ignore) --include-indels Do not skip indels -c CORES, --cores CORES Number of cores to use (default: all) vcf_files VCF files with genotypes
This is just a simple wrapper for samtools mpileup
usage: cisVar.py mpileup [-h] [-F PREFIX_NAME] [-r TRIAL_DEPTHS] -f ALLCHRFASTA -B SORTEDBAM [-p MPILEUPBEDFILE] Run mpileup optional arguments: -h, --help show this help message and exit Run Options: -F PREFIX_NAME, --SampleName PREFIX_NAME sample/population name (default: cis_var) -r TRIAL_DEPTHS, --readDepth TRIAL_DEPTHS minimum read depth per variant (default: 20) mpileup Options: -f ALLCHRFASTA, --fasta ALLCHRFASTA fasta file with all chromosomes (Required) -B SORTEDBAM, --BAMfile SORTEDBAM sorted BAM file (Required) -p MPILEUPBEDFILE, --mpileupBEDfile MPILEUPBEDFILE BED to use instead of the BED generated in the prep phase (Do not use if possible, use prep with limit instead)
This step actually calculates the POST-frequencies for the data.
usage: cisVar.py post [-h] [-F PREFIX_NAME] [-r TRIAL_DEPTHS] [-a GENOSFILE] Run POST frequency calculation optional arguments: -h, --help show this help message and exit Run Options: -F PREFIX_NAME, --SampleName PREFIX_NAME sample/population name (default: cis_var) -r TRIAL_DEPTHS, --readDepth TRIAL_DEPTHS minimum read depth per variant (default: 20) POST Options (Deprecated): -a GENOSFILE, --allelesFile GENOSFILE The genotypes file, (Optional, default is file created in prep)
This step converts the genotype file made in the prep step into a matrix that can be used in the regression. It is important that this genotype file is perfectly sorted to match the outputs of the POST step.
usage: cisVar.py geno [-h] [-F PREFIX_NAME] [-r TRIAL_DEPTHS] [-g GENOSFILE] [-i INDIVIDUALSLIST] Munge genotypes to prepare for regression optional arguments: -h, --help show this help message and exit Run Options: -F PREFIX_NAME, --SampleName PREFIX_NAME sample/population name (default: cis_var) -r TRIAL_DEPTHS, --readDepth TRIAL_DEPTHS minimum read depth per variant (default: 20) Genotype Options: -g GENOSFILE, --genoFile GENOSFILE The genotypes file, (Optional, default is file created in prep) -i INDIVIDUALSLIST, --individualsFile INDIVIDUALSLIST list of individuals matching genotype matrix; one indv per line
This is the actual regression step, it makes sure all the files are in the right
place and then calls regression_qtls.R
to do the actual regression.
usage: cisVar.py qtls [-h] [-F PREFIX_NAME] [-r TRIAL_DEPTHS] [-n NUMINDV] Run the regression optional arguments: -h, --help show this help message and exit Run Options: -F PREFIX_NAME, --SampleName PREFIX_NAME sample/population name (default: cis_var) -r TRIAL_DEPTHS, --readDepth TRIAL_DEPTHS minimum read depth per variant (default: 20) Regression Options: -n NUMINDV, --numberIndividuals NUMINDV The number of individuals in the pool (if omitted, calculated from genotype file length)
The regression produces z-scores and p-values, and additionally writes coefficients and some simple summary plots in separate files.
This step calls the open/closed alleles and produces a final integrated file with all available data as both a tad-delimited file and as a pandas dataframe.
usage: cisVar.py tidy [-h] [-F PREFIX_NAME] [-r TRIAL_DEPTHS] [-b BEDFILE] [-t TEXTFILE] [-p PANDASFILE] Tidy up regression, call open/closed optional arguments: -h, --help show this help message and exit Run Options: -F PREFIX_NAME, --SampleName PREFIX_NAME sample/population name (default: cis_var) -r TRIAL_DEPTHS, --readDepth TRIAL_DEPTHS minimum read depth per variant (default: 20) inputs: -b BEDFILE, --bedfile BEDFILE BED file to extract rsIDs from (optional) outputs: -t TEXTFILE, --textfile TEXTFILE Parsed output -p PANDASFILE, --pandasfile PANDASFILE Parsed dataframe
This option just downloads the Snakefile and config files from this repo, for easy access when code is installed via pip.
usage: cisVar.py get_snake [-h] [-x] Download Snakefile and config to current dir optional arguments: -h, --help show this help message and exit -x, --extra Get additional sample and cluster configs
This script is separate and is in the scripts
folder. It takes a search string
as an input and produces both combined and merged DataFrames. The combined
dataframe is just all dataframes combined in order with sample data added as a
column and to the index. The merged dataframe is a collapsed dataframe that has
one entry per SNP with p-values combined using Fisher's method and supporting
population data. It also includes information on the level of support for the
open and closed calls.
The search string should match your prefix and depth from the main pipeline. For
example, if you used a name of 'cis_var' plus a sample name (the variable part)
of e.g. CEU and YRI, and a read depth of 20, your search string would be:
cis_var.{sample}.20.regression.pd
.
The script will write cis_var.combined.20.regression.pd
and
cis_var.merged.20.regression.pd
.
usage: combine.py [-h] [-c COLUMN_NAME] [--no-merge] search_str Combine a bunch of cisVar pandas files by sample (e.g. population). Requires a search string such as prefix.{sample}.regression.pd. Writes ------ prefix.combined.regression.pd A simple merger of all DataFrames prefix.merged.regression.pd A per-snp merger based on p-value positional arguments: search_str e.g. name.{sample}.regression.pd, used to find files optional arguments: -h, --help show this help message and exit -c COLUMN_NAME, --column-name COLUMN_NAME Name for combine column, e.g. population --no-merge Produce only a combined dataframe, not a merged dataframe. merging can add half an hour over combination, which takes seconds
There is an additional script in scripts
called plot_fun.R
that takes a
single argument—the output of the regression step (e.g.
cis_var.YRI.20.totals.txt
) and creates a simple density pre-freq vs post freq
plot.