This is the repository for an EM algorithm of Wright-Fisher model for analyzing Evolve and Resequence (ER) data. From ER data, which contain allele frequency paths during experimental evolution for numerous SNPs, this program can estimate population genetics parameters i.e.) effective population size, selection coefficients and dominance parameters for each SNP or all SNPs.
We note that EMWER can be run by python2.7.
Please, check and install the required python libraries,
using pip.
$ pip install scipy numpy subprocess32
Please, check configuration using the following command, and resolve the dependencies. We note that EMWER depends on C++14 and OpenMP.
$ ./waf configure
Please, install using the following command
$ ./waf build
The executable file will be generated as build/default/WF_main
Although you can execute the above executable file directly, we recommend executing our program through python script EMWER.py as below:
$ python EMWER.py -c condition/emwer_test.cond -s data/real_data_test.sync
The -s specifies the synchronized file (see detail in https://sourceforge.net/p/popoolation2/wiki/Tutorial/) of ER data. The -c specifies the file name of the condition file. The -o specifies the file name of the estimation result.
The estimation result for two SNPs is like below:
250 0.0364121 0.5 0.000328661 0 3.84428 -1
250 0.0130407 0.5 0.000610262 0 0.278235 -1
Col1,2,3,4,5,6,7 respectively represent effective population size, selection coefficient, dominance parameter, half width of 95% confidence interval for estimated selection coefficient, that for dominance parameter, likelihood ratio, original sign of selection coefficients for variant allele. We noticed that, as default, selection coefficients and dominance parameter is those of the allele which is positively selected. Hence, if you want to derive these parameter for variant allele, you have to specify the condition allele as variant in condition file discussed below.
Experimental and estimation conditions can be specified in condition file as below:
genl:0,0,0,15,15,23,27,37,37,37
repl:1,4,5,4,5,1,5,4,5,1
pop:200
slc:0
dom:0.5
opt:s
disc:100
rbd:0.05,0.95
allele:positive
Each row can specify each experimental condition.;
genl specifies the list of generation when each column in the synchronized file was sampled.
repl specifies the list of replicate index of each column in the synchronized.
pop, slc and dom specifies initial values of estimation for effective population size, selection coefficient and dominance parameter, respectively.
disc specifies the maximum number of discrete states of allele frequency used in estimation.
opt specifies the first letter of parameter name which you want to optimize (p, s and d means effective population sizes, selection coefficients and dominance parameter respectively; you can conduct simultaneous estimation for selection and dominance by specifying opt as sd).
rbd specifies the upper and lower limit of allele frequency of population at 0th generation.
allele specifies the allele which selection coefficients and dominance parameters are defined for. When you specify allele as positive, EMWER calculate the parameters for positively selected alleles, which results in positive values for estimated selection coefficients. When you specify allele as variant, EMWER calculate the parameters for variant alleles.
This repository includes the code of three external libraries.
- Eigen is a C++ library for linear algebra. The code is included here.
- cmdline is a command line parser, which is implemented in C++. The code is included here.
- L-BFGS-B is a limited-memory quasi-Newton code for bound-constrained optimization, which is implemented in Fortran. We used it after converting it into C code by f2c. The code is included here.
EMWER is licensed under GLP Version 3. External libraries used in this repository are licensed as below:
- Eigen is primarily licensed under MPL2, although some part of third party code are licensed under BSD or LGPL licenses. Please, see the original license description for detail.
- cmdline is licensed under Modified BSD.
- L-BFGS-B is licensed under New BSD License.