PCAngsd

From software
Jump to: navigation, search

This page contains information about the program PCAngsd, which estimates the covariance matrix for low depth NGS data in an iterative procedure based on genotype likelihoods. Based on the population structure inference PCAngsd is able to estimate individual allele frequencies. By incorporating these allele frequencies in Empirical Bayes approaches, PCAngsd can perform PCA (estimate covariance matrix), call genotypes, estimate inbreeding coefficients (per-individual and per-site) and perform a genome selection scan using principal components in structured populations. The entire program is written in Python 2.7.

Pcangsd admix.gif
Simulated low depth NGS data of 3 populations


Download

The program can be downloaded from Github: https://github.com/Rosemeis/pcangsd

Latest release of PCAngsd: 0.3

git clone https://github.com/Rosemeis/pcangsd.git;
cd pcangsd/

The following Python packages are needed to run PCAngsd (found in all popular distributions): numpy, scipy and pandas.

PCAngsd should work on all platforms meeting the requirements but server-side usage is recommended.


Quick start

# See all options in PCAngsd
python pcangsd.py -h

# Only estimate covariance matrix 
python pcangsd.py -beagle test.beagle.gz -o test

# Estimate covariance matrix and inbreeding coefficients
python pcangsd.py -beagle test.beagle.gz -inbreed 1 -o test

# Estimate covariance matrix and perform selection scan
python pcangsd.py -beagle test.beagle.gz -selection 1 -o test

Input

The only input PCAngsd needs and accepts are genotype likelihoods in Beagle format. ANGSD can be easily be used to compute genotype likelihoods and output them in the required Beagle format.

./angsd -GL 1 -out genoLikes -nThreads 10 -doGlf 2 -doMajorMinor 1  -doMaf 2 -SNP_pval 1e-6 -bam bam.filelist

See ANGSD for more info on how to compute the genotype likelihoods and call SNPs.

Using PCAngsd

All the different options in PCAngsd are listed here. Usually all the desired analyses must be run in the same command, however PCAngsd can also be run in chunk-mode where per-site estimations are performed on a chunk of the data at a time using a pre-estimated covariance matrix. More information of chunk-mode estimations can be found here.

PCAngsd will always compute the covariance matrix (unless performing in chunk-mode estimations). It uses the computed principal components to estimate individual allele frequencies in an iterative procedure. This procedure is performed until the individual allele frequencies have converged.

-beagle [Beagle filename]

Path to file of the genotype likelihoods in Beagle format.

-beaglelist [filelist]

Parse a file with a list of multiple Beagle files, e.g. if the genotype likelihoods have been computed separately for each chromosome.

-M [int]

Maximum number of iterations for covariance estimation. Only needed in rare cases. (Default: 100)

-M_tole [float]

Tolerance value for the iterative covariance matrix estimation. (Default: 1e-4)

-EM [int]

Maximum number of EM iterations for computing the population allele frequencies. (Default: 200)

-EM_tole [float]

Tolerance value in EM algorithm for population allele frequencies estimation. (Default: 1e-4)

-e [int]

Manually select the number of eigenvalues to use in the modelling of individual allele frequencies. (Default: Automatically tested)

-reg [float]

Add regularization term in the modelling of individual allele frequencies to perform ridge regression. May help on convergence for individual allele frequencies. Must be used when scaling principal components prior to the modelling of individual allele frequencies.

-scaled

Scale significant principal components in relation to the top principal component using their corresponding eigenvalues prior to modelling individual allele frequencies.

-o [prefix]

Set the prefix for all output files created by PCAngsd (Default: "pcangsd").

LD can also be taken into account when computing the covariance matrix. LD regression has been implemented in PCAngsd.

-LD [int]

Select the number of preceding sites to use in LD regression.

Call genotypes

Genotypes can be called from posterior genotype probabilities incorporating the individual allele frequencies in prior.

-geno [float]

Call genotypes with defined threshold.

-genoInbreed [float]

Call genotypes with defined threshold also taking inbreeding into account. -inbreed [int] is required.

Inbreeding

Per-individual inbreeding coefficients incorporating population structure can be computed using three different methods:

-inbreed 1

A maximum likelihood estimator computed by an EM algorithm. Only allows for F-values between 0 and 1. Based on [1].

-inbreed 2

Simple estimator also computed by an EM algorithm. Based on ngsF.

-inbreed 3

(Not recommended for low depth NGS data!) Estimator using the kinship matrix. Based on PC-Relate.

-inbreed_iter [int]

Maximum number of iterations for the EM algorithm methods. (Default: 200)

-inbreed_tole [float]

Tolerance value for the EM algorithms for inbreeding coefficients estimation. (Default: 1e-4)


Per-site inbreeding coefficients incorporating population structure alongside likehood ratio tests for HWE can be computed as follows:

-inbreedSites

Selection

A genome selection scan can be computed using two different methods based on posterior expectations of the genotypes:

-selection 1

Using an extended model of FastPCA. Performs a genome selection scan along all significant PCs.

-selection 2

Using an extended model of PCAdapt.

Relatedness

Work in progress...

Estimate kinship matrix based on method Based on PC-Relate:

-kinship

Automatically estimated if -inbreed 3 has been selected.

Chunk-mode estimations

PCAngsd can also be run in chunk-mode, where a chunk of the data is processed at a time. This means that estimations on very large data sets are feasible for per-site parameters. In order to run PCAngsd in chunk-mode a pre-estimated covariance matrix must be provided. The estimation of the covariance matrix can be feasible by estimating it from a representative subset of the data set. Chunk-mode estimations are enabled by specifying the amount of sites to evaluate at a time:

-chunksize [int]

Number of sites to read in at a time for chunk-mode estimations.

-cov [file]

Covariance matrix file needed in order to perform chunk-mode estimations.

The following estimations can be performed in chunk-mode (individual allele frequencies are estimated and saved for all sites automatically):

-selection 1
-selection 2
-inbreedSites
-geno [float]

Note: Genotypes can also be called incorporating both individual allele frequencies and inbreeding coefficients, however one must also provide pre-estimated per-individual inbreeding coefficients as done with the covariance matrix:

-F [file]
-genoInbreed [float]

Citation