RelateAdmix

From software
Revision as of 10:06, 4 April 2016 by Albrecht (talk | contribs) (Run example using C)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Brief description

This page contains information about the program called relateAdmix, which can be used to infer relatedness coefficients for pairs of individuals even if they are admixed. The program has both an R interface and a C interface. Below is a description of how to install and use each of them. To be able to infer the relatedness you will need to know the individuals admixture proportions and the allele frequencies in each of the possible populations. This can be done e.g. using the program ADMIXTURE as shown in the example of how to use the C interface.

Installation - Linux/MAC

Both the C and R versions of the program have been made for linux but should work on other unix like systems. You need to have zlib installed. Some MAC users might also have to use makevars


Windows is not supported

Download

Download folder

Installation of R package

wget http://www.popgen.dk/software/download/relateAdmix/relateAdmix_0.14.tar.gz
R CMD INSTALL relateAdmix_0.14.tar.gz

MAC user might have to use curl instead of wget

Problems with zlib

If your R installation for some reason does not link to your zlib (e.i. gzopen missing) or you have a really old zlib installation then you can install the R part using

tar -xf relateAdmix_0.14.tar.gz
rm relateAdmix/src/Cinterface.*
rm relateAdmix/src/Makefile #only needed if you compiled the C program
R CMD INSTALL relateAdmix

Installation of C program

wget http://www.popgen.dk/software/download/relateAdmix/relateAdmix_0.14.tar.gz
tar -xvzf relateAdmix_0.14.tar.gz 
cd relateAdmix/src/ 
mv CPP_Makefile Makefile
make

Run example

Run example using R

After installing the package you can load it into R and try the example

library(relateAdmix)
example(relate)

This shows an example of how to use the package. More information can be found in the man pages

?relate


Run example using C

After installing the program you can try running it on the example data set in the data folder, which consists of 50 individuals that are admixed from 2 source populations.

If you are in the src folder where you installed relateAdmix and you have the software ADMIXTURE installed this can be done as follows:

cd ../data

# First run Admixture using a plink ".bed" as input to produce population specific allele 
# frequencies (smallPlink.2.P)  and individual ancestry proportions (smallPlink.2.Q).
# (note other programs can be used instead of Admixture, e.g. Structure and FRAPPE)
admixture smallPlink.bed 2 

# Then run relateAdmix with plink bed, bim and fam files plus the Admixture output files as input
../src/relateAdmix -plink smallPlink -f smallPlink.2.P -q smallPlink.2.Q -P 20

NB!. Only use binary plink (.bed) since ADMIXTURE switches allele frequencies when using .ped files


Quick R pdf plot from the commandline:

# Plot the results in R (R needs to be installed)
Rscript -e "r<-read.table('output.k',head=T,as.is=T);pdf('rel.pdf');plot(r[,4],r[,5],ylab='k2',xlab='k1');dev.off()"

RelAd.png


Output file format

Example of output

ind1    ind2    k0      k1      k2      nIter
0       1       0.999941        0.000038        0.000021        26
0       2       0.999979        0.000010        0.000011        29
0       3       0.999953        0.000029        0.000018        26
0       4       0.999952        0.000023        0.000025        26
0       5       0.999972        0.000020        0.000007        26
0       6       0.999995        0.000003        0.000002        26
0       7       0.999995        0.000003        0.000002        26
0       8       0.999894        0.000069        0.000038        32
0       9       0.999894        0.000069        0.000038        32
0       10      0.999903        0.000071        0.000026        26
0       11      0.999903        0.000071        0.000026        26


The first two columns are the individuals number. The next three columns are the estimated relatedness coefficients and the last column is the number of iterations used.


Input file format

The input consists of three files describignt the genotype data, a file with admixture proportions for each individual and a file with allele frequencies for each SNP for each source population. The genotype data files are plink bed/bim/fam files. And the remaining two files are in the output format for the program ADMIXTURE:

Example of the content of an admixture proportion file (for 3 populations)

0.531631 0.468359 0.000010
0.564461 0.435529 0.000010
0.850660 0.149330 0.000010
0.630527 0.369463 0.000010
0.747429 0.219346 0.033225
0.999980 0.000010 0.000010
0.999980 0.000010 0.000010
0.682072 0.317918 0.000010
0.000010 0.999980 0.000010
0.793133 0.206857 0.000010

Each row is an individual and each column is a population. The admixture proportions for each individual must sum to 1

Example of the allele frequency file (for 3 populations)

0.312722 0.208605 0.999990
0.881352 0.999990 0.966966
0.708206 0.838869 0.932119
0.427789 0.620694 0.532966
0.411998 0.622253 0.534072
0.427789 0.620694 0.532966
0.440817 0.581630 0.618751
0.733733 0.985281 0.953523
0.724083 0.451452 0.784607
0.811161 0.578612 0.787782

Each row is an SNP and each column is a population. When using plink files the allele frequency is the MAJOR allele frequency.

Citing and references

relateAdmix

Moltke, I, Albrechtsen, A (2013). RelateAdmix: a software tool for estimating relatedness between admixed individuals. Bioinformatics. pubmed bibtex

ADMIXTURE

D.H. Alexander, J. Novembre, and K. Lange. Fast model-based estimation of ancestry in unrelated individuals. Genome Research, 19:1655–1664, 2009.


change log

  • 0.14 made more MAC usable (I think). Thanks to Paul Lott for reporting it and for suggestions and Thorfinn Sand for changing it
  • 0.13 added extra check for file exists to give instant errors + changes all printf to fprintf(stderr,
  • 0.11 changed threading to a fixed pool of threads
  • 0.10 optimized code
  • 0.09 added error for when the number of sites and individuals does not match between files
  • 0.08 fixed a bug that would sometimes print an extra line when multiple threaded
  • 0.07 fixed a small leak