Skip to content

Samtools

Learning outcomes

After having completed this chapter you will be able to:

  • Use samtools flagstat to get general statistics on the flags stored in a sam/bam file
  • Use samtools view to:
    • compress a sam file into a bam file
    • filter on sam flags
    • count alignments
    • filter out a region
  • Use samtools sort to sort an alignment file based on coordinate
  • Use samtools index to create an index of a sorted sam/bam file
  • Use the pipe (|) symbol to pipe alignments directly to samtools to perform sorting and filtering

Material

Exercises

Alignment statistics

Exercise: Check out the statistics of the E. coli alignment by using samtools flagstat. Find the documentation here. Anything that draws your attention?

Answer

Code:

cd ~/workdir/alignment_output/
samtools flagstat SRR519926.sam

resulting in:

631808 + 0 in total (QC-passed reads + QC-failed reads)
0 + 0 secondary
0 + 0 supplementary
0 + 0 duplicates
627753 + 0 mapped (99.36% : N/A)
631808 + 0 paired in sequencing
315904 + 0 read1
315904 + 0 read2
302430 + 0 properly paired (47.87% : N/A)
624508 + 0 with itself and mate mapped
3245 + 0 singletons (0.51% : N/A)
0 + 0 with mate mapped to a different chr
0 + 0 with mate mapped to a different chr (mapQ>=5)

Of the reads, 47.87% is properly paired. The rest isn’t. Proper pairing is quite hard to interpret. It usually means that the 0x2 flag (each segment properly aligned according to the aligner) is false. In this case it means that the insert size is high for a lot of sequences. That is because the insert size distribution is very wide. You can find info on insert size distribution like this:

samtools stats SRR519926.sam | grep ^SN | cut -f 2,3

Now look at insert size average and insert size standard deviation. You can see the standard deviation is higher than the average, suggesting a wide distribution.

Compression, sorting and indexing

The command samtools view is very versatile. It takes an alignment file and writes a filtered or processed alignment to the output. You can for example use it to compress your SAM file into a BAM file. Let’s start with that.

Exercise: Create a script called 08_compress_sort.sh. Add a samtools view command to compress our SAM file into a BAM file and include the header in the output. For this, use the -b and -h options. Find the required documentation here. How much was the disk space reduced by compressing the file?

Tip: Samtools writes to stdout

By default, samtools writes it’s output to stdout. This means that you need to redirect your output to a file with > or use the the output option -o.

Answer

08_compress_sort.sh
#!/usr/bin/env bash

cd ~/workdir/alignment_output

samtools view -bh SRR519926.sam > SRR519926.bam
By using ls -lh, you can find out that SRR519926.sam has a size of 264 Mb, while SRR519926.bam is only 77 Mb.

To look up specific alignments, it is convenient to have your alignment file indexed. An indexing can be compared to a kind of ‘phonebook’ of your sequence alignment file. Indexing is done with samtools as well, but it first needs to be sorted on coordinate (i.e. the alignment location). You can do it like this:

samtools sort SRR519926.bam > SRR519926.sorted.bam
samtools index SRR519926.sorted.bam

Exercise: Add these lines to 08_compress_sort.sh, and re-run te script in order to generate the sorted bam file. After that checkout the headers of the unsorted bam file (SRR519926.bam) and the sorted bam file (SRR519926.sorted.bam) with samtools view -H. What are the differences?

Answer

Your script should like like this:

08_compress_sort.sh
#!/usr/bin/env bash

cd ~/workdir/alignment_output

samtools view -bh SRR519926.sam > SRR519926.bam
samtools sort SRR519926.bam > SRR519926.sorted.bam
samtools index SRR519926.sorted.bam

samtools view -H SRR519926.bam returns:

@HD     VN:1.0  SO:unsorted
@SQ     SN:U00096.3     LN:4641652
@PG     ID:bowtie2      PN:bowtie2      VN:2.4.2        CL:"/opt/conda/envs/ngs-tools/bin/bowtie2-align-s --wrapper basic-0 -x /config/workdir/ref_genome//ecoli-strK12-MG1655.fasta -1 /config/workdir/trimmed_data/paired_trimmed_SRR519926_1.fastq -2 /config/workdir/trimmed_data/paired_trimmed_SRR519926_2.fastq"
@PG     ID:samtools     PN:samtools     PP:bowtie2      VN:1.12 CL:samtools view -bh SRR519926.sam
@PG     ID:samtools.1   PN:samtools     PP:samtools     VN:1.12 CL:samtools view -H SRR519926.bam

And samtools view -H SRR519926.sorted.bam returns:

@HD     VN:1.0  SO:coordinate
@SQ     SN:U00096.3     LN:4641652
@PG     ID:bowtie2      PN:bowtie2      VN:2.4.2        CL:"/opt/conda/envs/ngs-tools/bin/bowtie2-align-s --wrapper basic-0 -x /config/workdir/ref_genome//ecoli-strK12-MG1655.fasta -1 /config/workdir/trimmed_data/paired_trimmed_SRR519926_1.fastq -2 /config/workdir/trimmed_data/paired_trimmed_SRR519926_2.fastq"
@PG     ID:samtools     PN:samtools     PP:bowtie2      VN:1.12 CL:samtools view -bh SRR519926.sam
@PG     ID:samtools.1   PN:samtools     PP:samtools     VN:1.12 CL:samtools sort SRR519926.bam
@PG     ID:samtools.2   PN:samtools     PP:samtools.1   VN:1.12 CL:samtools view -H SRR519926.sorted.bam

There are two main differences:

  • The SO tag at @HD type code has changed from unsorted to coordinate.
  • A line with the @PG type code for the sorting was added.

Note that the command to view the header (samtools -H) is also added to the header for both runs.

Filtering

With samtools view you can easily filter your alignment file based on flags. One thing that might be sensible to do at some point is to filter out unmapped reads.

Exercise: Check out the flag that you would need to filter for mapped reads. It’s at page 7 of the SAM documentation.

Answer

You will need the 0x4 flag.

Filtering against unmapped reads (leaving only mapped reads) with samtools view would look like this:

samtools view -bh -F 0x4 SRR519926.sorted.bam > SRR519926.sorted.mapped.bam

or:

samtools view -bh -F 4 SRR519926.sorted.bam > SRR519926.sorted.mapped.bam

Exercise: Generate a script called 09_extract_unmapped.sh to get only the unmapped reads (so the opposite of the example). How many reads are in there? Is that the same as what we expect based on the output of samtools flagstat?

Tip

Check out the -f and -c options of samtools view

Answer

Your script 09_extract_unmapped.sh should look like this:

09_extract_unmapped.sh
#!/usr/bin/env bash

cd ~/workdir/alignment_output

samtools view -bh -f 0x4 SRR519926.sorted.bam > SRR519926.sorted.unmapped.bam

Counting like this:

samtools view -c SRR519926.sorted.unmapped.bam

This should correspond to the output of samtools flagstat (631808 - 627753 = 4055)

samtools view also enables you to filter alignments in a specific region. This can be convenient if you don’t want to work with huge alignment files and if you’re only interested in alignments in a particular region. Region filtering only works for sorted and indexed alignment files.

Exercise: Generate a script called 10_extract_region.sh to filter our sorted and indexed BAM file for the region between 2000 and 2500 kb, and output it as a BAM file with a header.

Tip: Specifying a region

Our E. coli genome has only one chromosome, because only one line starts with > in the fasta file

cd ~/workdir/ref_genome
grep ">" ecoli-strK12-MG1655.fasta

gives:

>U00096.3 Escherichia coli str. K-12 substr. MG1655, complete genome

The part after the first space in the title is cut off for the alignment reference. So the code for specifying a region would be: U00096.3:START-END

Answer
10_extract_region.sh
#!/usr/bin/env bash

cd ~/workdir/alignment_output

samtools view -bh \
SRR519926.sorted.bam \
U00096.3:2000000-2500000 \
> SRR519926.sorted.region.bam

Redirection

Samtools is easy to use in a pipe. In this case you can replace the input file with a -. For example, you can sort and compress the output of your alignment software in a pipe like this:

my_alignment_command \
| samtools sort - \
| samtools view -bh - \
> alignment.bam

The use of -

In the modern versions of samtools, the use of - is not needed for most cases, so without an input file it reads from stdin. However, if you’re not sure, it’s better to be safe than sorry.

Exercise: Write a script called 11_align_sort.sh that maps the reads with bowtie2 (see chapter 2 of read alignment), sorts them, and outputs them as a BAM file with a header.

Answer
11_align_sort.sh
#!/usr/bin/env bash

TRIMMED_DIR=~/workdir/trimmed_data
REFERENCE_DIR=~/workdir/ref_genome
ALIGNED_DIR=~/workdir/alignment_output

bowtie2 \
-x $REFERENCE_DIR/ecoli-strK12-MG1655.fasta \
-1 $TRIMMED_DIR/paired_trimmed_SRR519926_1.fastq \
-2 $TRIMMED_DIR/paired_trimmed_SRR519926_2.fastq \
| samtools sort - \
| samtools view -bh - \
> $ALIGNED_DIR/SRR519926.sorted.mapped.frompipe.bam