vg pack error

[sdws1983]

1. What were you trying to do?

Run vg giraffe and vg pack to generate .pack file.

2. What did you want to happen?

Generate .pack file without errors

3. What actually happened?

Error occurred.

My command:

vg giraffe -t 40 -p -Z ref.giraffe.gbz -m ref.min -d ref.dist -f test_1.fq.gz -f test_2.fq.gz |  vg pack --threads 24 -x ref.giraffe.gbz -g - -o test.pack -Q 5

My logs:

Mapped 442147310 reads across 40 threads in 212111 seconds with 12.9819 additional single-threaded seconds.
Mapping speed: 52.1125 reads per second per thread
Used 1.34949e+06 CPU-seconds (including output).
Achieved 327.64 reads per CPU-second (including output)
Memory footprint: 39.0596 GB
vg: src/packer.cpp:883: size_t vg::Packer::average_node_quality(size_t) const: Assertion `total_node_quality(i) == 0' failed.
━━━━━━━━━━━━━━━━━━━━
Crash report for vg v1.60.0 "Annicco"
Stack trace (most recent call last) in thread 24550:
#10   Object "", at 0xffffffffffffffff, in 
#9    Object "/public2/home/software/vg", at 0x21cb233, in __clone
#8    Object "/public2/home/software/vg", at 0x212476a, in start_thread
#7    Object "/public2/home/software/vg", at 0x20c6ffd, in gomp_thread_start
#6    Object "/public2/home/software/vg", at 0x1296cd0, in vg::Packer::make_compact() [clone ._omp_fn.6]
#5    Object "/public2/home/software/vg", at 0x12965df, in vg::Packer::average_node_quality(unsigned long) const
#4    Object "/public2/home/software/vg", at 0x20f2f65, in __assert_fail
#3    Object "/public2/home/software/vg", at 0x5f201b, in __assert_fail_base.cold
#2    Object "/public2/home/software/vg", at 0x5f20f3, in abort
#1    Object "/public2/home/software/vg", at 0x20f9575, in raise
#0    Object "/public2/home/software/vg", at 0x21260cc, in __pthread_kill
ERROR: Signal 6 occurred. VG has crashed. Visit https://github.com/vgteam/vg/issues/new/choose to report a bug.
Please include this entire error log in your bug report!
━━━━━━━━━━━━━━━━━━━━

4. If you got a line like Stack trace path: /somewhere/on/your/computer/stacktrace.txt, please copy-paste the contents of that file here:

Place stacktrace here.

5. What data and command can the vg dev team use to make the problem happen?

6. What does running vg version say?

Place vg version output here

[ryandkuster]

Just to chime in for further context, I'm experiencing the exact same error at the vg pack step (vg v1.61.0 "Plodio") and can't seem to find a workaround. Giraffe/pack/call works for 94% of the samples, so it would be surprising if there were something wrong with the indexing or chopping of the .gbz used by vg giraffe or vg pack. This seems independent of the sample, and in some instances, identical samples split across multiple sequencing lanes will yield this vg pack error in only one technical replicate. Investigating those instances where two sequencing replicates produce gams that pass and fail the vg pack step, vg validate yields invalid alignments in both taking these two forms:

• "Edit erroneously claims match on node ___ between node position ___ and edit 0, position 0 on reverse strand"
• "Length of node ___ (n) exceeded by Mapping with offset n and from-length n:"

These invalid alignment warnings are found in both the passing and failing replicates. The graph is valid.

For additional context, I've constructed the pangenome as follows:

1. PGGB for chromosome-wise pangenome construction
2. ODGI squeeze to combine .og chromosomes
3. ODGI view to convert to .gfa
4. vg autoindex -X RG --workflow giraffe on the .gfa
5. vg giraffe -Z .gbz -m .min -d .dist -f R1.fq -f R2.fq
6. vg pack -x .rg -g .gam -Q 5 (also attempted 'vg pack -x .gbz -g .gam -Q 5` with same error)

Please let me know if there are any troubleshooting steps that would be helpful to assist with. The genotyping from the 94% of samples that did work is very promising!

I appreciate y'all!

[glennhickey]

Those validate messages aren't warnings, they are fatal errors denoting gbz/gam incompatiblility. So it would seem something went very wrong regardless of whether or not pack crashes.

Usually this is a result of using a different graph file between vg giraffe and vg pack. But if that's not the case here, then it would seem to be a bug in giraffe causing an incorrect mapping in the GAM.

If it is a giraffe bug, the only way forward might be to share some data to reproduce so one of the developers can debug it...

[ryandkuster]

@glennhickey I'm able to share the relevant files. Let me know the best contact and method of transfer. Thanks!

[glennhickey]

whatever's easiest -- maybe a google drive link or something.

[glennhickey]

@adamnovak @xchang1 @jltsiren

I've put some (private) data here, with a command line that runs giraffe to make an invalid GAM. Seems like a bug!

/private/home/ghickey/dev/work/kuster/REPRODUCE

[adamnovak]

I think the problem here is that this graph has cyclic end-to-start self edges on short nodes, so you can repeat them several times, and this is somehow confusing the code that splices tail alignments on, and getting it to put the wrong offsets in the mappings. We've never tested Giraffe at scale on graphs with these tight cycles, so we haven't caught this before.

The workaround would be to not have the graph look like that, but we should be able to have a proper fix for this soon.