ANGSD: Analysis of next generation Sequencing Data

Latest tar.gz version is (0.938/0.939 on github), see Change_log for changes, and download it here.

Abbababa2: Difference between revisions

From angsd
Jump to navigation Jump to search
No edit summary
No edit summary
 
(26 intermediate revisions by 2 users not shown)
Line 1: Line 1:
The ''ABBABABA (multipop)'' compute the abbababa test (aka D-statistic), that means testing for an ancient admixture (or wrong tree topology).
The ''ABBABABA (multipop)'' compute the abbababa test (aka D-statistic), that means testing for an ancient admixture (or wrong tree topology).
Differently from ABBABABA (D_stat) multiple individuals for each one of the groups are allowed. As all methods in ANGSD we require that the header of the BAM files are the same.
Differently from ABBABABA (D_stat) multiple individuals for each one of the groups are allowed. As all methods in ANGSD we require that the header of the BAM files are the same.
This method [http://www.g3journal.org/content/8/2/551 has a publication].
; some of the options only works for the developmental version availeble from github
; some of the options only works for the developmental version availeble from github
; if you use -rf to specify regions. These MUST appear in the same ordering as your fai file.
; if you use -rf to specify regions. These MUST appear in the same ordering as your fai file.
Line 8: Line 11:
  [*.bam and/or *.cram| NGS genome datasets{bg:orange}]->[Sequence data|All bases or Random bases]
  [*.bam and/or *.cram| NGS genome datasets{bg:orange}]->[Sequence data|All bases or Random bases]
[Sequence data]->[Elaborate multiple genomes per population]
[Sequence data]->[Elaborate multiple genomes per population]
[Sequence data]->[*.abbababa2counts|ABBA and BABA intermediate counts file {bg:blue}]
[Elaborate multiple genomes per population]->[*.abbababa2|weighted ABBA and BABA counts file {bg:blue}]
[Elaborate multiple genomes per population]->[*.abbababa2|weighted ABBA and BABA counts file {bg:blue}]
</classdiagram>
</classdiagram>


<classdiagram type="dir:LR">
<classdiagram type="dir:LR">
[*.abbababa2|weighted ABBA and BABA counts file {bg:blue}]->estAvgError.R[*.Observed.txt|Observed D stat and Z scores{bg:blue}]
[*.abbababa2|weighted ABBA and BABA counts file {bg:blue}]->estAvgError.R[D stat and Z scores{bg:blue}]
[*.abbababa2|weighted ABBA and BABA counts file {bg:blue}]->estAvgError.R[*.ErrorCorr.txt|D stat and Z scores Error Corrected{bg:blue}]
[*.abbababa2|weighted ABBA and BABA counts file {bg:blue}]->estAvgError.R[*.TransRemErrorCorr.txt|D stat and Z scores Error Corrected with ancient Transition Removal{bg:blue}]
[*.abbababa2|weighted ABBA and BABA counts file {bg:blue}]->estAvgError.R[*.RemTrans.txt|D stat and Z scores with Ancient Transition Removal{bg:blue}]
</classdiagram>
</classdiagram>


Line 35: Line 34:
-sample         0 sample a single base in each individual
-sample         0 sample a single base in each individual
-maxDepth         100 max depth of each site allowed
-maxDepth         100 max depth of each site allowed
-sizeH1         1 num of individuals in group H1
-sizeFile       (null)  file with sizes of the populations
-sizeH2         1 num of individuals in group H2
-sizeH3         1 num of individuals in group H3
-sizeH4         1 num of individuals in group H4
-enhance 0 only analyze sites where outgroup H4 is non poly
-enhance 0 only analyze sites where outgroup H4 is non poly
-Aanc         0 set H4 outgroup allele as A in each site
-Aanc         0 set H4 outgroup allele as A in each site
-combFile         0 create an optional *.abbababa2counts file where are printed the  
        -useLast                        0       1=use the last group of bam files as outgroup
                                    numbers of alleles combinations without having weighted the individuals
 
</pre>
</pre>
This function will counts the number of ABBA and BABA sites.
This function will counts the number of ABBA and BABA sites of the 4-population trees that can be built from the data, where the outgroup is fixed.
=Output=
;1)*.abbbababa2 (used for the 4-population test)
Each line represents a block with a chromsome name (Column 1), a start position (Column 2), an end postion (Column 3). Columns 4 and 5 are the numerator and denominator of the D-statistic for their specific block. Column 6 is the number of sites containing data in that block. The other 256 columns are the normalized counts of the 256 allele patterns between the 4 populations, starting from X0000=AAAA,X0001=AAAC,....,X3333=TTTT, with the correspondence 0=A,1=C,2=G,3=T. Every block is repeated a number of times corresponding to the trees that are built.
This file is used as input for the R script estAvgError.R.


=Options=
=Options=
Line 50: Line 50:
take all bases at each position.
take all bases at each position.
;-rmTrans [int]
;-rmTrans [int]
0; use all reads (default), 1 Remove transitions (important for ancient DNA)
0; use all reads (default), 1 Remove ancient transitions (important for ancient DNA)
;-blockSize [INT]
;-blockSize [int]
Size of each block. Choose a number that is higher than the LD in the populations. For human 5Mb (5000000) is usually used.  
Size of each block. Choose a number that is higher than the LD in the populations. For human 5Mb (5000000) is usually used.  
; -anc [fileName.fa]
; -anc [fileName.fa]
Include an outgroup in fasta format.
Include an outgroup in fasta format.
; -useLast [int]
1: use the last group of bam files as outgroup for the D-stat analysys. Default: 0 (use the fasta file as outgroup)
; -doCounts 1
; -doCounts 1
use -doCounts 1 in order to count the bases at each sites after filters.
use -doCounts 1 in order to count the bases at each sites after filters.
; -enhance [int]
; -enhance [int]
1: use only sites where the reads for the outgroup has the same base for all reads (If you want to use it, it must be only when you have one genome in the outgroup, it won't work otherwise).
1: use only sites where the reads for the outgroup has the same base for all reads.
; -sample [int]
; -sample [int]
1: sample only one base at each position for every individual 0: all bases at each position are used for the ABBABABA test
1: sample only one base at each position for every individual 0: all bases at each position are used for the ABBABABA test
; -maxDepth [int]
; -maxDepth [int]
allows for a maximum depth in each site to avoid overflow of the ABBA BABA counts
allows for a maximum depth in each site to avoid overflow of the ABBA BABA counts. Default 100.
; -sizeH* [int]
; -sizeFile [fileName]
decide how many individuals are in each group (the file list must contain the BAM files ordered from population 1 to 4).  
file that specifies number of individuals in each population (more than 4 populations can be defined). If not provided, it is assumed that each population has only one individual.
If you are using a fasta file (option -anc) for population H4, leave -sizeH4 at its default value
; -Aanc [int]
; -Aanc [int]
1: H4 allele is A in each site.
1: H4 allele is A in each site.
; -combFile [int]
1: create an intermediate *.abbababa2counts to obtain the allele events before weighting the samples (however, this file is not used for the estimation of the D-statistic).


In order to do fancy filtering of bases based on quality scores see the [[Alleles_counts|Allele counts]] options.
In order to do fancy filtering of bases based on quality scores see the [[Alleles_counts|Allele counts]] options.


=Tutorial of the ABBABABA (Multipop) test=
=Tutorial for the ABBABABA (Multipop) test=
 
This tutorial require having Samtools previously installed, and the library 'pracma' previously installed in R.
;Some preparation steps before using ANGSD
== Prepare BAM and FASTA files ==
 
Download the latest version of angsd in your working folder from the github repository
Download the latest version of angsd in your working folder from the github repository
<pre>
<pre>
Line 84: Line 82:
<pre>
<pre>
ln -s ./angsd/angsd ANGSD
ln -s ./angsd/angsd ANGSD
ln -s ./angsd/R/estAvgError.R RSCRIPT
ln -s ./angsd/R/estAvgError.R DSTAT
</pre>
</pre>
Get 10 example .bam datasets, position them in the folder ./bams/ and create a file bam.filelist listing the pathnames of those datasets  
Get 10 example .bam datasets, position them in the folder ./bams/ and create a file bam.filelist listing the pathnames of those datasets  
Line 97: Line 95:
<pre>
<pre>
cat bam.filelist
cat bam.filelist
 
</pre>
<pre>
bams/smallNA06985.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
bams/smallNA06985.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
bams/smallNA06994.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
bams/smallNA06994.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
Line 108: Line 107:
bams/smallNA11832.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
bams/smallNA11832.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
bams/smallNA11840.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
bams/smallNA11840.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
<\pre>
</pre>
Download a fasta file for the chimpanzee. This is going to be used as the outgroup for the four-population test. One can use a bam file as well (see in one of the other examples after the tutorial how to do it).
Download a fasta file for the chimpanzee. This is going to be used as the outgroup for the 4-population test.
<pre>
<pre>
wget http://popgen.dk/software/download/angsd/hg19ancNoChr.fa.gz
wget http://popgen.dk/software/download/angsd/hg19ancNoChr.fa.gz
Line 115: Line 114:
gunzip chimpHg19.fa.gz
gunzip chimpHg19.fa.gz
samtools faidx chimpHg19.fa #indexing the fasta file
samtools faidx chimpHg19.fa #indexing the fasta file
rm chimpHg19.fa.gz
</pre>
</pre>
Now, generate a fasta file for one of our 10 bam file. We assume such a genome has very high quality and we can use it as a reference for estimating error rates in others of our datasets.
Now, generate a fasta file for one of our 10 bam file. We assume such a genome has very high quality and we can use it as a reference for estimating error rates in others of our datasets.
<pre>
<pre>
./ANGSD -i bams/smallNA11840.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam -doFasta 1 -out perfectSampleCEU
./ANGSD -i bams/smallNA11840.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam -doFasta 1 -doCounts 1 -out perfectSampleCEU
gunzip perfectSampleCEU.fa.gz
gunzip perfectSampleCEU.fa.gz
samtools faidx perfectSampleCEU.fa
samtools faidx perfectSampleCEU.fa
</pre>
</pre>
In this tutorial we perform the ABBABABA test on H1,H2,H3,H4 consisting respectively of 3,5,2,1 individuals. In this case we use a fasta file for the outgroup H4 (we use a bam file in next tutorial). We will apply error correction to groups H1 and H2 assuming one of the samples from H3 as high-quality reference one.


;Prepare files for the estimation of type specific error rates
==Generate files for the error correction==
We will apply error correction to the group with 3 individuals, using "perfectSampleCEU" as high-quality reference genome.
The population containing 3 individuals affected by transition error goes from line 6 to line 8 in the file bam.filelist. We select those individuals and write them in another file.
<pre>
sed -n 6,8p bam.filelist > bamWithErrors.filelist
</pre>
and then we use "doAncError" to generate the intermediate files that we will use later as input for the R script that calculates the D-statistic. "doAncError" applies the so called "perfect individual assumption", based on which error rates are estimate using a high quality genome (option -ref) and an outgroup (option -anc), both in fasta format. We have already prepared the two fasta files in our preparation phase.
<pre>
./ANGSD -doAncError 1 -anc chimpHg19.fa -ref perfectSampleCEU.fa -out errorFile -bam bamWithErrors.filelist
</pre>


Assume population H1 consists of the first two genomes of our list, while population H2 consists of the genomes 3 to 7. We want to apply error correction to those genomes, because we know they have been subjected to contamination. We generate two files containing the pathnames of the genomes of H1 and H2 on which we want to apply error correction.
==4-population test==
In this tutorial we perform the ABBABABA test on all the combinations of 4 populations amongst 6 populations of size 1,2,2,3,2,1 individuals, where the last population is fixed as outgroup (so that there are 30 possible combinations). The last population is represented by the fasta file defined with the option -anc, of which we enable the use as outgroup by the option -useLast 0. One can use the last population of .bam files as outgroup with the option -useLast 1. Create a file named sizeFile.size and write the size of each population (IT IS NECESSARY to define the size of the -anc outgroup population, that is always 1):
<pre>
<pre>
sed -n 1,2p bam.filelist > bamH1.filelist
1
sed -n 3,7p bam.filelist > bamH2.filelist
2
2
3
2
1
</pre>
</pre>
and then we use "doAncError" to generate the intermediate files that we will use later to estimate the error rates for the two groups H1 and H2. "doAncError" apply the so called "perfect individual assumption", based on which error rates are estimate using a high quality genome (option -ref) and an outgroup (option -anc), both in fasta format. We have already prepared the two fasta files in our preparation phase.
We decide to target three chromosomes, one of the three with loci between position 10Mb and 15Mb. Thus create a file called regions.txt in which is written
<pre>
<pre>
./ANGSD -doAncError 1 -anc chimpHg19.fa -ref perfectSampleCEU.fa -out bamH1 -bam bamH1.filelist
1:
./ANGSD -doAncError 1 -anc chimpHg19.fa -ref perfectSampleCEU.fa -out bamH2 -bam bamH2.filelist
5:
16:10000000-15000000
</pre>
</pre>
The output of ANGSD will show no data about chromosome 1. This happens when all blocks within that chromosome contained no data and therefore where not printed.


;ABBABABA test
After running ANGSD to count ABBA and BABA combinations, we will call the R script who applies error correction to the ABBA and BABA allele combinations and produces the final output files.
 
Now, we want to run the four population test using:
H1: first 2 bam files
H2: bam files from 3 to 7
H3: bam files from 8 to 10
H4: chimpHg19.fa file
After running ANGSD we will call the R script who apply error correction to the ABBA and BABA allele combinations and produce the final output files.
<pre>
<pre>
./ANGSD -doAbbababa2 1 -bam bam.filelist -doCounts 1 -out bam.AllelePatterns -sizeH1 2 -sizeH2 5 -sizeH3 3 -anc chimpHg19.fa -minQ 20 -minMapQ 30
./ANGSD -doAbbababa2 1 -bam bam.filelist -sizeFile sizeFile.size -doCounts 1 -out bam.Angsd -anc chimpHg19.fa -rf regions.txt -useLast 0 -minQ 20 -minMapQ 30 -p 1
</pre>
</pre>
The output file is
[[The output file is]]
;bam.AllelePatterns.abbbababa2 (used for the 4-population test)
[[bam.Angsd.abbbababa2 (used for the 4-population test)]]
Each line represents a block with a chromsome name (Column 1), a start position (Column 2), an end postion (Column 3). The new columns are the counts of all 256 counted combination of alleles, starting from X0000=AAAA,X0001=AAAC,....,X3333=TTTT, with the correspondence 0=A,1=C,2=G,3=T.
Each line represents a block with a chromsome name (Column 1) for one of the possible 30 trees (so each block is written on 30 lines), a start position (Column 2), an end postion (Column 3). Columns 4,5 and 6 are the numerator, denominator and number of sites analyzed in the block. The next 256 columns are the counted patterns of alleles in the tree, starting from X0000=AAAA,X0001=AAAC,....,X3333=TTTT, with the correspondence 0=A,1=C,2=G,3=T.
This file is used as input for the R script estAvgError.R.
This file is used as input for the R script estAvgError.R.
Optionally, one can also produce
;bam.AllelePatterns.abbbababa2counts (optional file)
As above each line represents a block with a chromsome name (Column 1), a start position (Column 2), an end postion (Column 3). The new columns are the counts of all 256 counted combination of alleles, starting from X0000=AAAA,X0001=AAAC,....,X3333=TTTT, with the correspondence 0=A,1=C,2=G,3=T. This file is not used as input for the ABBABABA test.


We run the R script specifying the intermediate error files for populations H1 and H2. We also want to study the effect of error correction if we add individually to each population an error rate between 0 and 0.005 with step 0.001 and involving transitions A->T and C-->T. It is also possible to specify the names of H1,H2,H3 to be seen on the plot. In this case we use the generic names CEU1,CEU2,CEU3. When at least an error file is given as input, the script will apply error correction.
We run the R script specifying the error files for the population with 3 individuals. This is done defining the error files in each populations inside a text file (including a line for the outgroup population). If a population has no error file, it is sufficient to write NA. Create a file called errorList.error with written
<pre>
<pre>
Rscript RSCRIPT angsdFile="bam.AllelePatterns" out="result" file1="bamH1.ancError" file2="bamH2.ancError"  addErr="0,0.005,0.001;A,C;T;CEU1,CEU2,CEU3"
NA
NA
NA
./errorFile.ancError
NA
NA
</pre>
</pre>
The script will show the calculated D statistic along with Z-score, Pvalues, Standard deviation and other quantities.
Create a file popNames.name with written
<pre>
<pre>
--- Table of Results ---
Population1
---------------------------------------------------------------------------------
Population2
  Mode |Dstat |sd(Dstat) |Djack |Zscore |Pvalue
Population3
---------------------------------------------------------------------------------
PopWithError
Observed |-6.323e-02 |6.985e-02 |-6.323e-02 |-0.905 |3.7e-01
Population4
---------------------------------------------------------------------------------
Chimpanzee
Err Corr |-6.430e-02 |7.226e-02 |-6.431e-02 |-0.890 |3.7e-01
---------------------------------------------------------------------------------
No Trans |-1.141e-02 |6.311e-02 |-1.141e-02 |-0.181 |8.6e-01
---------------------------------------------------------------------------------
Err Corr | | | | |
  and |-1.494e-02 |6.615e-02 |-1.496e-02 |-0.226 |8.2e-01
No Trans | | | | |
---------------------------------------------------------------------------------
plots with effect of removed errors and D statistic files for all the removed errors are in folder result.errorDataFolder
</pre>
</pre>
Those results are also contained in four distinct files
Run the Rscript with the command
;1) result.Observed.txt
D-statistic calculated WITHOUT Error Correction and WITHOUT Ancient Transition removal
<pre>
<pre>
mean(D) JK-D    V(JK-D) Z      pvalue  nABBA  nBABA  nBBAA
Rscript DSTAT angsdFile="bam.Angsd" out="result" sizeFile=sizeFile.size errFile=errorList.error nameFile=popNames.name
-0.063233      -0.063233      0.004878        -0.905320      0.365296        246.033565      279.248560      292.834879
</pre>
</pre>
;2) result.ErrorCorr.txt
The script will show the calculated D statistic along with Z-score, Pvalues, Standard deviation and other quantities for all 30 4-populations trees. Note: If error correction is not needed, it is sufficient to avoid specifying any error file. If no names need to be provided, the script will assign Population_* as standard name. If no size file is provided, the script assigns 1 to each population. At least one between the name file and the size file is needed. It is possible to recycle the size file used in ANGSD.
 
The D-statistics and other informations are contained in four distinct files depending on the application of error correction and ancient transition removal. The files are named as follow:
;[[1)result.Observed.txt]]
D-statistic calculated WITHOUT Error Correction and WITHOUT Ancient Transition removal
;[[2) result.ErrorCorr.txt]]
D-statistic calculated WITH Error Correction and WITHOUT Ancient Transition removal
D-statistic calculated WITH Error Correction and WITHOUT Ancient Transition removal
<pre>
;[[3) result.ErrorCorr.TransRem.txt]]
mean(D) JK-D    V(JK-D) Z      pvalue  nABBA  nBABA  nBBAA
-0.064295      -0.064309      0.005221        -0.889833      0.373555        238.242964      270.983960      293.326044
</pre>
;3) result.TransRemErrorCorr.txt
D-statistic calculated WITH Error Correction and WITH Ancient Transition removal
D-statistic calculated WITH Error Correction and WITH Ancient Transition removal
<pre>
;[[4) result.TransRem.txt]]
mean(D) JK-D    V(JK-D) Z      pvalue  nABBA  nBABA  nBBAA
-0.014939      -0.014959      0.004376        -0.225829      0.821335        81.636843      84.112983      293.326044
</pre>
;4) result.RemTrans.txt
D-statistic calculated WITHOUT Error Correction and WITH Ancient Transition removal
D-statistic calculated WITHOUT Error Correction and WITH Ancient Transition removal
Specifically, the values contained in the four files are: mean(D)=average D-stat, JK-D=jackknife estimate of the D-stat, V(JK-D)=variance of the D-stat, Z=Z score, pvalue=pvalue from the Z score, nABBA=number of ABBA patterns observed, nBABA=number of BABA patterns observed, nBlocks=number of blocks with observed data, H*=the names of the four populations for the specific tree. Note that the number of patterns might not be integer because of how ANGSD treats multiple genomes per populations.
=Cite the method=
<pre>
<pre>
mean(D) JK-D    V(JK-D) Z      pvalue  nABBA  nBABA  nBBAA
@article{Soraggi2018,
-0.011406      -0.011406      0.003983        -0.180730      0.856580        85.730478      87.708709      292.834879
author = {Soraggi, S. and Wiuf, C. and Albrechtsen, A.},
doi = {10.1534/g3.117.300192},
issn = {21601836},
journal = {G3: Genes, Genomes, Genetics},
number = {2},
title = {{Powerful inference with the D-statistic on low-coverage whole-genome data}},
volume = {8},
year = {2018}
}
</pre>
</pre>
Specifically, the values contained in the four files are: mean(D)=average D-stat, JK-D=jackknife estimate of the D-stat, V(JK-D)=variance of the D-stat, Z=Z score, pvalue=pvalue from the Z score, nABBA=number of ABBA patterns observed, nBABA=number of BABA patterns observed, nBBAA=all the other observed patterns. Note that the number of patterns might not be integer because of how ANGSD treats multiple genomes per populations.
In case of error correction, the R script also creates the folder result.errorDataFolder containing:
-the file barPlotErrors.pdf showing a barplot of the error rates (plot shown below)
-the file plotAddErr.A2T.pdf showing the effect of error correction on transition A-->T (plot shown below)
-the file errorRates.txt showing in each line transition errors for each population, respectively
-all the files related to the addition of error correction to H1,H2,H3, necessary to plot the files plotAddErr.A2T.pdf.
[[File:barPlotErrorJPGs.jpg]]
[[File:PlotAddErr.A2TJPG.jpg]]

Latest revision as of 10:34, 26 June 2018

The ABBABABA (multipop) compute the abbababa test (aka D-statistic), that means testing for an ancient admixture (or wrong tree topology). Differently from ABBABABA (D_stat) multiple individuals for each one of the groups are allowed. As all methods in ANGSD we require that the header of the BAM files are the same.

This method has a publication.

some of the options only works for the developmental version availeble from github
if you use -rf to specify regions. These MUST appear in the same ordering as your fai file.

<classdiagram type="dir:LR">

[*.bam and/or *.cram| NGS genome datasets{bg:orange}]->[Sequence data|All bases or Random bases]

[Sequence data]->[Elaborate multiple genomes per population] [Elaborate multiple genomes per population]->[*.abbababa2|weighted ABBA and BABA counts file {bg:blue}] </classdiagram>

<classdiagram type="dir:LR"> [*.abbababa2|weighted ABBA and BABA counts file {bg:blue}]->estAvgError.R[D stat and Z scores{bg:blue}] </classdiagram>

Method

Brief Overview

> ./angsd -doAbbababa2

--------------
abcDstat2.cpp:
	-doAbbababa2	                0	run the abbababa analysis
	-rmTrans		        0       remove transitions
	-blockSize		       5000000	size of each block in bases
	-anc			       (null)	fasta file with outgroup
	-sample			        0	sample a single base in each individual
	-maxDepth		        100	max depth of each site allowed
	-sizeFile		       (null)   file with sizes of the populations	
	-enhance			0	only analyze sites where outgroup H4 is non poly
	-Aanc			        0	set H4 outgroup allele as A in each site
        -useLast                        0       1=use the last group of bam files as outgroup

This function will counts the number of ABBA and BABA sites of the 4-population trees that can be built from the data, where the outgroup is fixed.

Output

1)*.abbbababa2 (used for the 4-population test)

Each line represents a block with a chromsome name (Column 1), a start position (Column 2), an end postion (Column 3). Columns 4 and 5 are the numerator and denominator of the D-statistic for their specific block. Column 6 is the number of sites containing data in that block. The other 256 columns are the normalized counts of the 256 allele patterns between the 4 populations, starting from X0000=AAAA,X0001=AAAC,....,X3333=TTTT, with the correspondence 0=A,1=C,2=G,3=T. Every block is repeated a number of times corresponding to the trees that are built. This file is used as input for the R script estAvgError.R.

Options

-doAbbababa2 1

take all bases at each position.

-rmTrans [int]

0; use all reads (default), 1 Remove ancient transitions (important for ancient DNA)

-blockSize [int]

Size of each block. Choose a number that is higher than the LD in the populations. For human 5Mb (5000000) is usually used.

-anc [fileName.fa]

Include an outgroup in fasta format.

-useLast [int]

1: use the last group of bam files as outgroup for the D-stat analysys. Default: 0 (use the fasta file as outgroup)

-doCounts 1

use -doCounts 1 in order to count the bases at each sites after filters.

-enhance [int]

1: use only sites where the reads for the outgroup has the same base for all reads.

-sample [int]

1: sample only one base at each position for every individual 0: all bases at each position are used for the ABBABABA test

-maxDepth [int]

allows for a maximum depth in each site to avoid overflow of the ABBA BABA counts. Default 100.

-sizeFile [fileName]

file that specifies number of individuals in each population (more than 4 populations can be defined). If not provided, it is assumed that each population has only one individual.

-Aanc [int]

1: H4 allele is A in each site.

In order to do fancy filtering of bases based on quality scores see the Allele counts options.

Tutorial for the ABBABABA (Multipop) test

This tutorial require having Samtools previously installed, and the library 'pracma' previously installed in R.

Prepare BAM and FASTA files

Download the latest version of angsd in your working folder from the github repository

https://github.com/ANGSD/angsd.git

Create symbolic links to angsd and the necessary R script

ln -s ./angsd/angsd ANGSD
ln -s ./angsd/R/estAvgError.R DSTAT

Get 10 example .bam datasets, position them in the folder ./bams/ and create a file bam.filelist listing the pathnames of those datasets

wget http://popgen.dk/software/download/angsd/bams.tar.gz
tar xf bams.tar.gz
for i in bams/*.bam;do samtools index $i;done #index bam files
ls bams/*.bam > bam.filelist
rm bams.tar.gz #remove zipped file

This is how the file bam.filelist looks like

cat bam.filelist
bams/smallNA06985.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
bams/smallNA06994.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
bams/smallNA07000.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
bams/smallNA07056.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
bams/smallNA07357.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
bams/smallNA11829.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
bams/smallNA11830.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
bams/smallNA11831.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
bams/smallNA11832.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam
bams/smallNA11840.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam

Download a fasta file for the chimpanzee. This is going to be used as the outgroup for the 4-population test.

wget http://popgen.dk/software/download/angsd/hg19ancNoChr.fa.gz
mv hg19ancNoChr.fa.gz chimpHg19.fa.gz
gunzip chimpHg19.fa.gz
samtools faidx chimpHg19.fa #indexing the fasta file

Now, generate a fasta file for one of our 10 bam file. We assume such a genome has very high quality and we can use it as a reference for estimating error rates in others of our datasets.

./ANGSD -i bams/smallNA11840.mapped.ILLUMINA.bwa.CEU.low_coverage.20111114.bam -doFasta 1 -doCounts 1 -out perfectSampleCEU
gunzip perfectSampleCEU.fa.gz
samtools faidx perfectSampleCEU.fa

Generate files for the error correction

We will apply error correction to the group with 3 individuals, using "perfectSampleCEU" as high-quality reference genome. The population containing 3 individuals affected by transition error goes from line 6 to line 8 in the file bam.filelist. We select those individuals and write them in another file.

sed -n 6,8p bam.filelist > bamWithErrors.filelist

and then we use "doAncError" to generate the intermediate files that we will use later as input for the R script that calculates the D-statistic. "doAncError" applies the so called "perfect individual assumption", based on which error rates are estimate using a high quality genome (option -ref) and an outgroup (option -anc), both in fasta format. We have already prepared the two fasta files in our preparation phase.

./ANGSD -doAncError 1 -anc chimpHg19.fa -ref perfectSampleCEU.fa -out errorFile -bam bamWithErrors.filelist

4-population test

In this tutorial we perform the ABBABABA test on all the combinations of 4 populations amongst 6 populations of size 1,2,2,3,2,1 individuals, where the last population is fixed as outgroup (so that there are 30 possible combinations). The last population is represented by the fasta file defined with the option -anc, of which we enable the use as outgroup by the option -useLast 0. One can use the last population of .bam files as outgroup with the option -useLast 1. Create a file named sizeFile.size and write the size of each population (IT IS NECESSARY to define the size of the -anc outgroup population, that is always 1):

1
2
2
3
2
1

We decide to target three chromosomes, one of the three with loci between position 10Mb and 15Mb. Thus create a file called regions.txt in which is written

1:
5:
16:10000000-15000000

The output of ANGSD will show no data about chromosome 1. This happens when all blocks within that chromosome contained no data and therefore where not printed.

After running ANGSD to count ABBA and BABA combinations, we will call the R script who applies error correction to the ABBA and BABA allele combinations and produces the final output files.

./ANGSD -doAbbababa2 1 -bam bam.filelist -sizeFile sizeFile.size -doCounts 1 -out bam.Angsd -anc chimpHg19.fa -rf regions.txt -useLast 0 -minQ 20 -minMapQ 30 -p 1

The output file is bam.Angsd.abbbababa2 (used for the 4-population test) Each line represents a block with a chromsome name (Column 1) for one of the possible 30 trees (so each block is written on 30 lines), a start position (Column 2), an end postion (Column 3). Columns 4,5 and 6 are the numerator, denominator and number of sites analyzed in the block. The next 256 columns are the counted patterns of alleles in the tree, starting from X0000=AAAA,X0001=AAAC,....,X3333=TTTT, with the correspondence 0=A,1=C,2=G,3=T. This file is used as input for the R script estAvgError.R.

We run the R script specifying the error files for the population with 3 individuals. This is done defining the error files in each populations inside a text file (including a line for the outgroup population). If a population has no error file, it is sufficient to write NA. Create a file called errorList.error with written

NA
NA
NA
./errorFile.ancError
NA
NA

Create a file popNames.name with written

Population1
Population2
Population3
PopWithError
Population4
Chimpanzee

Run the Rscript with the command

Rscript DSTAT angsdFile="bam.Angsd" out="result" sizeFile=sizeFile.size errFile=errorList.error nameFile=popNames.name

The script will show the calculated D statistic along with Z-score, Pvalues, Standard deviation and other quantities for all 30 4-populations trees. Note: If error correction is not needed, it is sufficient to avoid specifying any error file. If no names need to be provided, the script will assign Population_* as standard name. If no size file is provided, the script assigns 1 to each population. At least one between the name file and the size file is needed. It is possible to recycle the size file used in ANGSD.

The D-statistics and other informations are contained in four distinct files depending on the application of error correction and ancient transition removal. The files are named as follow:

1)result.Observed.txt

D-statistic calculated WITHOUT Error Correction and WITHOUT Ancient Transition removal

2) result.ErrorCorr.txt

D-statistic calculated WITH Error Correction and WITHOUT Ancient Transition removal

3) result.ErrorCorr.TransRem.txt

D-statistic calculated WITH Error Correction and WITH Ancient Transition removal

4) result.TransRem.txt

D-statistic calculated WITHOUT Error Correction and WITH Ancient Transition removal

Specifically, the values contained in the four files are: mean(D)=average D-stat, JK-D=jackknife estimate of the D-stat, V(JK-D)=variance of the D-stat, Z=Z score, pvalue=pvalue from the Z score, nABBA=number of ABBA patterns observed, nBABA=number of BABA patterns observed, nBlocks=number of blocks with observed data, H*=the names of the four populations for the specific tree. Note that the number of patterns might not be integer because of how ANGSD treats multiple genomes per populations.

Cite the method

@article{Soraggi2018,
author = {Soraggi, S. and Wiuf, C. and Albrechtsen, A.},
doi = {10.1534/g3.117.300192},
issn = {21601836},
journal = {G3: Genes, Genomes, Genetics},
number = {2},
title = {{Powerful inference with the D-statistic on low-coverage whole-genome data}},
volume = {8},
year = {2018}
}