Explanation of BAM compression levels

Question

[noodle]

Hi Biostars,

Can someone point me to documentation, or describe here, what the practical differences are for different compression levels of BAM files and why you might choose one over another? For example, in bedtools functions you can pass -ubam, or with picard you can pass --COMPRESSION_LEVEL, among others...

Answer 1

[LChart]

BAM files are block-gzip compressed (https://manpages.ubuntu.com/manpages/impish/man1/bgzip.1.html), so the "compression level" refers precisely to the gzip compression level used on each individual block (https://www.rootusers.com/gzip-vs-bzip2-vs-xz-performance-comparison/ for actual impact).

[noodle]

Thank you for pointing to these pages, do you know why you would use one compression level over another? For example, high compression for files that will be stored vs no compression for files regularly accessed? Is there any guidance for at what point (file size, access frequency, etc) different compression levels should be considered?

[LChart]

For archival purposes, you're better off using CRAM than BAM; and for individual bam files the difference in performance and space for various compression levels is, in my experience, negligible.