This small utility takes in a bam file and creates a pair of FASTQ files per read group.
I created this utility specifically to allow me to realign sequencing data from the Cancer Cell Line Encyclopedia. For that data, only bams are available and I needed to align them to a different build of the human genome to facilitate certain analyses. I tried several different approaches including Picard SamToFastq, Tophat's bam2fastx and Samtools -> unix shell tools. Each had problems which I found discouraging. For each of these approaches I found that producing the fastq files took longer than STAR to align them.
python bam2fastq.py [bam]
2 Fastq files per read group named [RG].1.fq and [RG].2.fq
This utility requires samtools to be installed and in the path. No unusual python modules are used. Memory requirements are small. In practice a maximum of 50 mb was seen. The program can take advantage of 2 cores, but should work with 1 core. Samtools and this program run concurrently, but in practice the python uses a full core while samtools uses 30-50% of a core. In my hands, on my desktop which uses an Intel i5-3470 processor I see about 1 million reads processed per 10 sec.
This little script allows you to generate unmapped SAM files from
mapped BAM files. This will let you directly remap mapped BAM files
to a new reference without a FASTQ intermediate. To use this with
STAR, add --readFileType SAM PE or SAM SE as appropriate and
"--readFilesCommand bam2sam.py" or "--readFilesCommand bam2sam.py
--se". This will remove all of the mapping information, flip the
reads if they are inverted when aligned, and strip out any mapping
flags other than read groups. In practice, I've seen mapping speeds
comparable with STAR on FASTQ files. Using this for on the fly
remapping will consume 1.5 cores and up to 200MB of memory in addition
to STAR's requirements.