Fetching, parsing, and analyzing data from the 1000 genomes project.
Python Shell
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


Welcome to the Mutation Sets Code!
(c) 2015 Genevieve Shattow

Last updated: 7 September 2015

There are several pieces to this repository, so please read the README
before science-ing.

NB: This code has been tested in a bash environment on a Linux server. I cannot
speak to its reliability on another system yet.

NB: The dataset used for development is 20130502 and has been hardcoded into the
code. If you want to use a different set, I hope it has the same format and 
directory structure on the server.

The purpose of this code is to fetch, parse, and analyze data from the 1000 
genomes project. Because the files are stored on the 1000genomes server, it 
does not make sense to store them locally as well (unless you have limited 
internet access). So this code fetches and unzips the files, extracts the 
necessary information, and gets rid of the local copy. It then cross-matches
the extracted data (which person has which mutations), with the mutation's 
sift score (how bad it is). Then it performs a few analyses, including 
plotting. A more detailed description and a how-to is below.

Code in the repository:
README (this document, right here)

Extra data in repository:
for definitions of each population, see: 

each txt file has a list of individual names/codes that are in that
population, with ALL including all of them.

To run:
Step 0:
Download repository, cd into code directory

Step 1:
./individuals.bash -c C

Step 2:

Step 3:
python mutation_sets.py -c C -pop POP or ./run_mutation_sets.sh

Step 4: 
Cure cancer

To use:
./individuals.bash -c C (each chromosome individually)


./run_individuals.sh (all chromosomes in sequence)
individuals.bash fetches and parses the phase 3 data from the 1000 genomes by 
chromosome. These files list all of the variants in that chromosome and which
individuals have each one. Each individual is tested twice. If both tests have 
values greater than 0, then the individual is said to have that variant. The 
indices of these variants are then copied into files for each individual. This 
has to be done for each chromosome, which can individually take a few hours to 
half a day (as run on an 8GB core). To run them all in succession, use 
run_individuals.sh. It takes no arguments and runs chromosomes 1-22.

For more specific information, this is at the top of the file:

# To use:
# ./individuals.bash -c N
# To separate out the mutation lists for each individual:
# step 0 - download each chromosome file and move the long names to a shorter name
#			e.g. ALL.chr1.phase3.......vcf.gz becomes ALL.chr1.individuals.vcf.gz
# step 1 - gunzip files
# step 2 - make new directories
# step 3 - run the oneliner grep -ve.... in 100 person batches, making sure the
#			chromosome number and file name are set.
#			This creates files of line numbers where individuals have mutations
#			on both alleles, corresponding to an rs number.
# 			nb - I've tried doing it in 1 go, it was going to take a millennium
#			nb2 - it should take .5-2 hours to run each 100, depending on the machine
#			and size of the chromosome
# step 4 - take the line numbers with mutations from the chrA.p files and rename them
#			into chrA.HGID in the chrAn directory. rm the chrA.p files that you don't 
# 			need anymore (they are really big)
# step 5 - rm the unzipped (really large) file

To use:

sifting.bash finds the sift scores of all of the variants, as calculated by the
Variant Effect Predictor (more information: www.ensembl.org/info/docs/tools/vep/).
The VEP has been run on the 1000 genomes data, but these files do not have the 
results linked to the individual patients. sifting.bash fetches the release data
for each chromosome. Not every variant has a sift score, so this script pulls out
the variants that do. It copies the variant ids (the index for this code set, rs
number, and within ENSEMBL), the sift score, and the phenotype.

From the top of the file:

# This script runs on the output of the VEP which includes the sift scores.
# For each chromosome, it moves and unzips the files.
# Then taking only the lines with sift scores, it records the line number 
# (minus the header) it came from.
# It looks up the ENSEMBL and HGNC gene cataplgues corresponding to each rs number
# using symbol_lookup.py. (optional)
# Then it combines all the columns into 
# line number, rs number, ENSEMBL ID, (HGNC ID), sift score, and phenotype score.

To use:
python mutation_sets.py -c C -pop POP


This script is the analysis/plotting part of this package. There are two 
sections after the individual files and variant sift scores are read in. 
The data is cut by sift scores less than a certain values, which is 0.05 as 
a default, but can be reset within the code.

The first section measures overlap in mutations among pairs of individuals by 
population. Considering two people, if both people have a mutation in a given 
variant then they are considered to have an overlap. Individuals are paired 
randomly without replacement, giving N/2 pairs, with N being the total number
of individuals. This analysis is run 100 times (hard coded in for now) to 
reduce any signal from accidental pairings between family members.

These results are then plotted as a histogram of the number of pairs (y-axis) 
with a given overlap (x-axis) in files named:
with C being the chromosome, sX.XX being the max sift score, and POP the 
population. The pairings are also output as text files named
with each line being N/2 numbers long, where each number is the overlaps in
that pair. There are 100 lines in the file - one for each run. 

The second analysis portion finds the number of individuals that have mutations
on a given pair of genes. This information is output in files named:
and are in the format of 
('gene1', 'gene2')	n
with gene1 and gene2 being the two genes in the pair and n being the number of
individuals that have mutations on both genes. This data can be used in 
conjunction with network data which weights each pair
(e.g. http://cagt10.bu.edu/blinghu/manuscript/Additional_File_1.txt.zip)

Fun code creating a d3js visualisation of co-mutated genes. It's not fully 
developed, just something I threw together. It uses the gene_pairs*.txt files
and one (EUR) is hard-coded in. The figure opens in a browser and allows the 
user to mouse over the heatmap with the values and gene pair appearing over 
each square.

Directory Structure
The github repository is downloaded as 

			Mutation Sets/
		|                   |
	  code/				populations/
After running the scripts, the directory structure looks like this:

		scripts described above
		dataset/ (e.g. 20130502)
				(temporary files from individuals.bash)
				chrC.HG* (created by individuals.bash)
				chrC.NA* (created by individuals.bash)
		Fdataset/ (e.g. F20130502)
			sifted.SIFT.chrC.txt (output from sifting.bash)
		individual_pairs_*.txt (output from mutation_sets.py)
		gene_pairs_*.txt  (output from mutation_sets.py)
		pair_distribution_*.png  (output from mutation_sets.py)

Or This:
|         			  |					|					 |					 |
code/				populations/	  output/				data/				plots/
|         			  |					|					 |					 |
scripts				1kg_POP.txt		individual_pairs_*		 |		pair_distribution_*
described							gene_pairs_*			 |					
above														 |
										    |								  |
											|								  |
										dataset/						   Fdataset/
										(e.g. 20130502)					(e.g. F20130502)
											|								|
									________|________				sifted.SIFT.chrC.txt
									|				|				(output from sifting.bash)
									chrCp/		chrCn/
									 |				|
						(folder	for 			chrC.HG*
						temporary output of		chrC.NA*
						individuals.bash)		(output from individuals.bash)
(best viewed with a screen > 95 characters wide)