注册 登录  
 加关注
   显示下一条  |  关闭
温馨提示!由于新浪微博认证机制调整,您的新浪微博帐号绑定已过期,请重新绑定!立即重新绑定新浪微博》  |  关闭

云之南

风声,雨声,读书声,声声入耳;家事,国事,天下事,事事关心

 
 
 

日志

 
 
关于我

专业背景:计算机科学 研究方向与兴趣: JavaEE-Web软件开发, 生物信息学, 数据挖掘与机器学习, 智能信息系统 目前工作: 基因组, 转录组, NGS高通量数据分析, 生物数据挖掘, 植物系统发育和比较进化基因组学

网易考拉推荐

Calling SNPs  

2016-09-30 09:37:03|  分类: 生信分析软件 |  标签: |举报 |字号 订阅

  下载LOFTER 我的照片书  |

http://ged.msu.edu/angus/tutorials-2011/snp_tutorial.html


http://popgen.dk/wiki/index.php/SNP_calling

Calling SNPs with Samtools

In this tutorial, we’re going to use the sorted BAM files we generated earlier to generate a list of SNPs at which the iso-female lines we re-sequenced differ from the reference Drosophila melanogaster genome. We’ll be using Samtools for this, but there is other software out there that will call SNPs and genotypes for you. We encourage you to try them out on your own.

Getting the data

Set up your EC2 server, and mount the provided snapshot on /data.

Next, install Samtools. When you are installing Samtools, make sure to install bcftools as well (which is included as part of the Samtools package):

%% cd /data
%% curl -O -L http://sourceforge.net/projects/samtools/files/samtools/0.1.16/samtools-0.1.16.tar.bz2
%% tar xvfj samtools-0.1.16.tar.bz2
%% cd samtools-0.1.16
%% make
%% cp samtools /usr/local/bin
%% cd misc/
%% cp *.pl maq2sam-long maq2sam-short md5fa md5sum-lite wgsim /usr/local/bin/
%% cd ..
%% cd bcftools
%% cp *.pl bcftools /usr/local/bin/

Since it took a while to do map our reads to the reference genome last time, we have the sorted bam files saved in the /data/snp_calling/ folder:

%% cd /data/snp_calling/

We’re going to try out two separate SNP callers, so let’s try to keep ourselves organized:

%% mkdir samtools_snps
%% cd samtools_snps

Calling SNPs with Samtools

To generate a BCF file (which is a binary data format used to hold information about sequence variants such as SNPs), run the following lines:

%% samtools mpileup -uD -r 2L:100,000-150,000 -f /data/drosophila/dmel-all-chromosome-r5.37.fasta ../RAL357_full_bwa.sorted.bam ../RAL391_full_bwa.sorted.bam | bcftools view -bvcg - > RAL_samtools.raw.bcf

You’ll notice that this command has two parts, and we’ve used several flags in the first part of it. Here’s what they mean: -u tells it to output into an uncompressed bcf file (rather than compressed) -D tells it to keep read depth for each sample -r tells it which chromosome region to call SNPs for (you can omit this if you want to do the whole genome, but in the interest of speed, we picked a 50kb region) -f tells it that the next argument is going to be the reference genome file

The Samtools portion of this calculates our genotype likelihoods. We then pipe the output to bcftools, which does our SNP calling based on those likelihoods. This portion of the command has several options as well. The -b flag tells it to output to BCF format (rather than VCF); -c tells it to do SNP calling, and -v tells it to only output potential variant sites (i.e., exclude monomorphic ones); and -g tells it to call genotypes for each sample in addition to just calling SNPs. Then we run:

%% bcftools view RAL_samtools.raw.bcf | vcfutils.pl varFilter -D100 > RAL_samtools.vcf

This line converts the BCF file into a VCF file (a flat text file rather than a binary, making it a lot easier to view), and then we pipe that into vcfutils.pl with the varFilter -D100 option, which filters out SNPs that had read depth higher than 100 (we don’t want to trust SNPs at sites with super high coverage, because they might be represent variation between variable copy number repeats, i.e., the reads that map to this location in the reference are actually from duplicated sites in your sample; you can–and should–change this parameter based on the kind of coverage you have in your dataset, e.g., -D500).

How does samtools detect SNPs? Every time a mapped read shows a mis-match from the reference genome, it does some fancy statistics to try and figure out whether the mis-match is because of a real SNP. It incorporates different types of information, such as the number of different reads that share a mis-match from the reference, the sequence quality data, and the expected sequencing error rates, and it essentially figures out whether it’s more likely that the observed mis-match is due simply to a sequencing error, or because of a true SNP.

The resulting VCF file has a lot of information about your genotypes. For a detailed description of the VCF format, see http://vcftools.sourceforge.net/specs.html.

Fow now, let’s look at the first few lines of the VCF file. The file lists the chromosome, position, ID, reference allele at that nucleotide position, and alternate alleles detected in our dataset (across all samples). It also tells you the “Quality” – which is basically a measure of how confident Samtools is that there really is a SNP there (higher is better) – and whether or not that SNP passed the quality filters. The “Info” field tells you various statistics about each position; the information in this field can vary, but what exactly each symbol (e.g., NS, DP, etc.) should be explained in the file header. The “Format” field tells you what type of data are found in the rest of the fields (there should be one additional field for each sample that you ran through Samtools). For example, GT:GQ:DP means that the sample fields for this position tell you the genotype, genotype quality, and depth (coverage) of each sample that you fed into Samtools; again, explanations of these symbol names (GT, GT, DP) can be found in the header of your VCF file.:

%% less RAL_samtools.vcf

Calling SNPs with GATK’s Unified Genotyper

Now let’s try mapping our reads with another mapper, the Unified Genotyper included as part of the Genome Analysis Toolkit (GATK). Again, let’s try to keep our files organized:

%% cd /data/snp_calling
%% mkdir GATK_snps
%% cd GATK_snps
%% curl -O ftp://ftp.broadinstitute.org/pub/gsa/GenomeAnalysisTK/GenomeAnalysisTK-latest.tar.bz2
%% tar xvfj GenomeAnalysisTK-latest.tar.bz2

Like FastQC, this is also written in java, so let’s make sure java is installed:

%% java -version

Now let’s run GATK’s Unified Genotyper. The GATK people recommend a few quality control steps before you run the SNP calling. In particular, they recommend local re-alignment around indels, because reads whose ends map to the location of an indel can sometimes lead to false positive SNP calls. I’ll try to illustrate this with a simplified example. Suppose the reference sequence is GGGGTTTT and there’s an alternate allele with a 4-bp insertion of C’s: GGGGCCCCTTTT. Now if a read carries the entire insertion, the aligner can figure things out without a problem. However, if the end of the read overlaps the insertion, you could run into problems. For example:

Reference : GGGG----TTTT
Read 1    :    GCCCCT
Read 2    : GGGG----C
                    *

Read 1 looks fine, but read 2 is probably mis-aligned; rather than counting the C as an insertion, which is probably better, the aligner has placed it after the indel, making it look like there is a SNP there. In GATK we can account for this problem by doing local re-alignment around potential indel sites, in which we incorporate information about all the reads in that region simultaneously, rather than mapping each read individually. First, GATK needs to figure out which regions are in need of re-alignment:

%% java -Xmx1g -jar GenomeAnalysisTK-1.0.5974/GenomeAnalysisTK.jar -T RealignerTargetCreator -R /data/drosophila/dmel-all-chromosome-r5.37.fasta -I ../RAL357_full_bwa.sorted.bam -o RAL357.realign.intervals -L 2L:100000-150000

And then it needs to actually do the re-alignment (this step is slow, taking ~10 min, even for our small region):

%% java -Xmx4g -jar GenomeAnalysisTK-1.0.5974/GenomeAnalysisTK.jar -I ../RAL357_full_bwa.sorted.bam -R /data/drosophila/dmel-all-chromosome-r5.37.fasta -T IndelRealigner -targetIntervals RAL357.realign.intervals -o ../RAL357_full_bwa.realigned.bam

Now we need to do this re-alignment again with the other fly strain that we genotyped (if you were running a large number of samples, this is where a shell script would come in handy). You might want to run this in the background or in a new session while you’re re-aligning the first one:

%% java -Xmx1g -jar GenomeAnalysisTK-1.0.5974/GenomeAnalysisTK.jar -T RealignerTargetCreator -R /data/drosophila/dmel-all-chromosome-r5.37.fasta -I ../RAL391_full_bwa.sorted.bam -o RAL391.realign.intervals -L 2L:100000-150000
%% java -Xmx4g -jar GenomeAnalysisTK-1.0.5974/GenomeAnalysisTK.jar -I ../RAL391_full_bwa.sorted.bam -R /data/drosophila/dmel-all-chromosome-r5.37.fasta -T IndelRealigner -targetIntervals RAL391.realign.intervals -o ../RAL391_full_bwa.realigned.bam

Now we would like to run the SNP caller, but we first need to use a package called Picard to fix a minor formatting problem in the re-aligned BAM files before GATK’s Unified Genotyper will accept them (these steps take a few minutes each as well):

%% cd /data/snp_calling/
%% curl -O -L http://sourceforge.net/projects/picard/files/picard-tools/1.47/picard-tools-1.47.zip
%% unzip picard-tools-1.47.zip
%% java -jar picard-tools-1.47/AddOrReplaceReadGroups.jar I= RAL357_full_bwa.realigned.bam O= RAL357_full_bwa.realigned.fixed.bam SORT_ORDER=coordinate RGID=RAL357 RGLB=RAL357 RGPL=illumina RGPU=RAL357 RGSM=RAL357 CREATE_INDEX=True VALIDATION_STRINGENCY=LENIENT
%% java -jar picard-tools-1.47/AddOrReplaceReadGroups.jar I= RAL391_full_bwa.realigned.bam O= RAL391_full_bwa.realigned.fixed.bam SORT_ORDER=coordinate RGID=RAL391 RGLB=RAL391 RGPL=illumina RGPU=RAL391 RGSM=RAL391 CREATE_INDEX=True VALIDATION_STRINGENCY=LENIENT

And finally we can run GATK’s SNP caller:

%% cd /data/snp_calling/GATK_snps
%% java -jar GenomeAnalysisTK-1.0.5974/GenomeAnalysisTK.jar -R /data/drosophila/dmel-all-chromosome-r5.37.fasta -T UnifiedGenotyper -I ../RAL357_full_bwa.realigned.fixed.bam -I ../RAL391_full_bwa.realigned.fixed.bam -o RAL_GATK.vcf -stand_call_conf 50.0 -stand_emit_conf 10.0 -dcov 500 -L  2L:100000-150000

These parameters are similar, but not identical, to those in Samtools. -stand_emit_conf 10.0 means that it won’t report any potential SNPs with a quality below 10.0; but unless they meet the quality threshold set by -stand_call_conf (50.0, in this case), they will be listed as failing the quality filter. -dcov 500 means that any site that has more than 500x coverage, the genotype caller will only use 500 randomly selected reads (for computational efficiency).

Keep in mind that, at this point, indel calling in GATK’s Unified Genotyper does not seem to be well supported. For that, you may want to stick with Samtools or other software for now.

Another approach you could take is to just do the local re-alignment with GATK, but then do your SNP calling using Samtools on the locally re-aligned BAM files.:

%% cd /data/snp_calling/samtools_snps
%% samtools mpileup -uD -r 2L:100,000-150,000 -f /data/drosophila/dmel-all-chromosome-r5.37.fasta ../RAL357_full_bwa.realigned.fixed.bam ../RAL391_full_bwa.realigned.fixed.bam | bcftools view -bvcg - > RAL_samtools_fixed.raw.bcf
%% bcftools view RAL_samtools_fixed.raw.bcf | vcfutils.pl varFilter -D100 > RAL_samtools_fixed.vcf

Comparing output

Now say you want to count how many SNPs each SNP calling approach found. You could do this pretty simply using grep on the vcf file. In this case, we only looked at a region on chromosome 2L, and the chromosome is the first thing listed about each SNP. So we can search for and then count every line in the vcf file that begins with the text ‘2L’. To use grep to search for text at the beginning of a line, you use the ^ symbol. Once we find the lines in the vcf file that describe our SNPs, we’re going to pipe them all to the wc command (with the -l flag) to count how many of them there are:

%% cd /data/snp_calling
%% grep '^2L' samtools_snps/RAL_samtools.vcf | wc -l
%% grep '^2L' GATK_snps/RAL_GATK.vcf | wc -l
%% grep '^2L' samtools_snps/RAL_samtools_fixed.vcf | wc -l

Which approach detected the most SNPs in this region?

There is a package called vcftools that has all sorts of utilities for working with VCF files. I won’t go over it here, but consult the project’s website and documentation at http://vcftools.sourceforge.net/ if you are interested. For now, compare the SNPs that each SNP caller detected by viewing the VCF files using less:

%% less samtools_snps/RAL_samtools.vcf
%% less GATK_snps/RAL_GATK.vcf
%% less samtools_snps/RAL_samtools_fixed.vcf

Another exercise:

In the snp_calling directory, you will also find BAM files generated by aligning the same set of reads to the same reference genome for one of the two fly lines (RAL357) using bowtie rather than bwa. Use Samtools to call SNPs and generate a VCF file on the bowtie alignment and compare it to the VCF file you got from the bwa alignment.

  评论这张
 
阅读(64)| 评论(0)
推荐 转载

历史上的今天

在LOFTER的更多文章

评论

<#--最新日志,群博日志--> <#--推荐日志--> <#--引用记录--> <#--博主推荐--> <#--随机阅读--> <#--首页推荐--> <#--历史上的今天--> <#--被推荐日志--> <#--上一篇,下一篇--> <#-- 热度 --> <#-- 网易新闻广告 --> <#--右边模块结构--> <#--评论模块结构--> <#--引用模块结构--> <#--博主发起的投票-->
 
 
 
 
 
 
 
 
 
 
 
 
 
 

页脚

网易公司版权所有 ©1997-2016