Which compression methods are suitable for use with cram-js?

[grabear]

This is mostly a repeat of another issue in the IGV repo. I was wondering if you would take a look:

igvteam/igv.js#2078

Is there anything that can be done to read CRAM files that have these (newer?) compression methods?

gzip
gzip-min
r32x16-o0
r32x16-o0P
r32x16-o1
r32x16-o1P
r32x16-o1R
r4x16-o0
r4x16-o0P
r4x16-o0R
r4x16-o1
r4x16-o1P
r4x16-o1R
raw
rNx16-cat
tok3-rans

[cmdcolin]

Hi there,
I am open to getting support for any compression values that appear in real world crams, and certainly open to PR particularly if it has test data.

We have also made a bunch of changes to this repo in past ~months to add more wasm parsing, leveraging cross compiled hts-codecs C code to wasm, not sure if igv.js had caught up with these changes, but even i don't know the comprehensive list of codecs available, so would probably be worth doing an audit

[cmdcolin]

(totally missed this issue when first posted, apologies about that!)

[grabear]

Hi @cmdcolin. No worries. I don't have the knowledge to update this codebase (C+JS), but I do know samtools pins the htscodecs here: https://github.com/samtools/htslib (https://github.com/samtools/htscodecs/tree/master)

If there's anything else I can do to help please let me know.

[cmdcolin]

I suspect this was related to #141 which indeed as you had noted in the igv.js thread is related to when CRAM v3.1 codecs were enabled by default in samtools. I believe this should be fixed in latest igv.js and jbrowse now but if you still see issues please let me know.

I did a large scale scan (AI assisted) scan of the codecs we implement in this repo and cross referenced current samtools, htslib, hts-specs documentation and found very few things that we do not support. see https://github.com/GMOD/cram-js/blob/main/CODEC_SUPPORT.md

The only things it flagged were

⚠️ rANS 32x16
⚠️ Striped variants - not exposed

Results https://github.com/GMOD/cram-js/blob/main/CODEC_SUPPORT.md grain of salt being ai generated but probably fairly accurate

[cmdcolin]

It additionally said that there was not even really possible to manually trigger those codecs, could be worth implementing but likely very rare to see in practice