Skip to content

Instantly share code, notes, and snippets.

What would you like to do?
Extract paired FASTQ reads from a BAM mapping file

Extracting paired FASTQ read data from a BAM mapping file

Sometimes FASTQ data is aligned to a reference and stored as a BAM file, instead of the normal FASTQ read files. This is okay, because it is possible to recreate raw FASTQ files based on the BAM file. The following outlines this process. The useful software samtools and bedtools are both required.

From each bam, we need to extract:

  1. reads that mapped properly as pairs
  2. reads that didn’t map properly as pairs (both didn’t map, or one didn’t map)

For #1, the following command will work. This was taken from this webpage.

samtools view -u -f 1 -F 12 > lib_002_map_map.bam

The -f and -F filter using flags in column 2 of the BAM file. These aren't always intuitive, and I won't describe them more here, but you can use this handy tool to better understand. Also note that the -u flag creates uncompressed BAM output rather than default compressed BAM output, so the files will be larger. This helps with quicker reading in later steps, but it isn't necessary to include this if you want to save disk space. samtools is super fast either way.

Resolving #2 is more complicated, as there are three ways a read might not have mapped as a proper pair. A. The first read mapped but the paired read did not. B. The first read did not map but the paired read did. C. Neither paired read mapped at all. Again, flags will be used to filter the original BAM file. This information was found at this webpage.

# R1 unmapped, R2 mapped
samtools view -u -f 4 -F 264 > lib_002_unmap_map.bam
# R1 mapped, R2 unmapped
samtools view -u -f 8 -F 260 > lib_002_map_unmap.bam
# R1 & R2 unmapped
samtools view -u -f 12 -F 256 > lib_002_unmap_unmap.bam

As you might expect, you have to then merge the three files that contain at least one unmapped pair.

samtools merge -u lib_002_unmapped.bam lib_002_unmap_map.bam lib_002_map_unmap.bam lib_002_unmap_unmap.bam

Next, these BAM files must be resorted so that they are ordered by read ID instead of location in the reference.

samtools sort -n lib_002_map_map.bam lib_002_mapped.sort
samtools sort -n lib_002_unmapped.bam lib_002_unmapped.sort

At this time, it is a good idea to check that you have the correct number of reads and no redundancy. You can summarize the original BAM file to get an idea of where you started.

samtools flagstat
## output
# 160673944 + 0 in total (QC-passed reads + QC-failed reads)
# 44648692 + 0 duplicates
# 136861465 + 0 mapped (85.18%:-nan%)
# 160673944 + 0 paired in sequencing
# 80336972 + 0 read1
# 80336972 + 0 read2
# 123173914 + 0 properly paired (76.66%:-nan%)
# 123173914 + 0 with itself and mate mapped
# 13687551 + 0 singletons (8.52%:-nan%)
# 37581548 + 0 with mate mapped to a different chr
# 28422566 + 0 with mate mapped to a different chr (mapQ>=5)

Notice the toal number of input reads that is found on the first line. You want to be sure that the number of unmapped and mapped reads total this number. It is easy to check using the following commands.

samtools view -c lib_002_mapped.sort.bam
## output
# 123173914
samtools view -c lib_002_unmapped.sort.bam
## output
# 37500030

Note that one paired read is counted as two reads here. If you sum these two numbers, they should equal the number you noted above, as they do here.

If all is good, you can now extract the FASTQ reads into two paired read files, as follows.

bamToFastq -i lib_002_mapped.sort.bam -fq lib_002_mapped.1.fastq -fq2 lib_002_mapped.2.fastq 
bamToFastq -i lib_002_unmapped.sort.bam -fq lib_002_unmapped.1.fastq -fq2 lib_002_unmapped.2.fastq 

And then it also makes sense to combine both the first and paired reads together from the mapped and unmapped files.

cat lib_002_mapped.1.fastq lib_002_unmapped.1.fastq > lib_002.1.fastq
cat lib_002_mapped.2.fastq lib_002_unmapped.2.fastq > lib_002.2.fastq

These two files should now have the same number of reads that are exactly as you would have received them if they had come directly from the sequencer as FASTQ.

Please also note that all of the commands above can be piped together in bash using |, which will save on disk space and time. So it is best to combine commands where possible.


This comment has been minimized.

Copy link

@splaisan splaisan commented Apr 29, 2018

Thanks for this nice tutorial, but does samtools view fastq not do most of it already?

What about doing it from a full genome BAM but only extract read pairs for which at least one read of a pair maps in a region?
I am preparing this for a training where I want students to map quickly so I thought of keeping a nice depth of coverage but a width for only a region or a single shorter chromosome.

From what I find, I need to pass one first time and collect the ID's (read names) of read that map in the region for at least one of the pair, then pass a second time through the BAM/SAM with tools like grep to collect all reads having these IDs/names and save to file. After which I could reformat the read properly in fastq pairs.
I fear I will have multimapping reads that will complicate the story and introduce duplicates (using uniq applied to fastq after concatenation may help but will cost extra time!).

Any suggestions or corrections on my plan?



This comment has been minimized.

Copy link

@ipstone ipstone commented Aug 21, 2018

Here is an useful link for a related topic:

The approach listed above, won't extract the matching pair sequence to the unmapped sequence read; for that purpose, I first collect the unmapped sequence, and then filter from the orginal bam files - all the Ids matching the 'unmapped', to obtain all the paired reads (map-unmap, unmap-map, unmap-unmap all together).


This comment has been minimized.

Copy link

@cfc424 cfc424 commented Jul 20, 2019

this is awsome!


This comment has been minimized.

Copy link

@jelber2 jelber2 commented Jul 25, 2019

Should have -o for newer versions of samtools (at least 1.9)

samtools sort -n lib_002_map_map.bam lib_002_mapped.sort -> samtools sort -n lib_002_map_map.bam -o lib_002_mapped.sort.bam
samtools sort -n lib_002_unmapped.bam lib_002_unmapped.sort -> samtools sort -n lib_002_unmapped.bam -o lib_002_unmapped.sort.bam

This comment has been minimized.

Copy link

@jelber2 jelber2 commented Jul 25, 2019

Should also mention that you might need to run [BBTools]( []( to repair the read1 and read2 pairing in=lib_002.1.fastq in2=lib_002.2.fastq out=lib_002.1.fastq.gz out2=lib_002.2.fastq.gz repair

This comment has been minimized.

Copy link

@hauduc hauduc commented Sep 20, 2020

I believe you need to add -h to the samtools view commands to include headers so samtools merge reads in the bams correctly, as well as -b if you want to save on storage space, not just removing the -u tag.


This comment has been minimized.

Copy link

@arumds arumds commented Feb 3, 2021

bam2fastq also outputs unpaired mapped and unpaired unmapped reads. How do we include or concat them to generate paired-end read files?


This comment has been minimized.

Copy link

@JoseAMontero JoseAMontero commented Jul 30, 2021

good tutorial

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment