Fail to assembly de-novo genome

[Dv1t]

Hello VStrains team, thank you for developing such a great tool, but while using it, I faced the following problem.
I attempted to assemble the complete HIV genome from this sample: SRR29407826. I used corona-spades and it worked fine.
However VStrains crashes when assembling it.
First, there was an error related to rev_dict in VStrains_PE_Inference.py. It doesn't had lowercase nucleotides in it and therefore raised KeyError. I fixed it replacing:
rev_dict = {"A": "T", "T": "A", "C": "G", "G": "C"}
with this:
rev_dict = { "A": "T", "T": "A", "C": "G", "G": "C", "a": "t", "t": "a", "c": "g", "g": "c" }
But new issue occurred, after messages in CLI log:

----------------------Paired-End Information Alignment----------------------
Start aligning reads to gfa nodes
Number of processed reads: 0

It freezes forever and do not proceed any further.

Worth mentioning details
In the same log there is a suspicious message:

INFO - graph kmer size: 0

Also VStrains can't read assembly_graph_after_simplification.gfa file (which is the output of spades) without changing its version in header from 1.2 to 1.0 manually.

Steps to reproduce

1. Assembly with spades:

spades.py --corona -1 SRR29407826_1.fastq -2 SRR29407826_2.fastq -o spades_G_SRR29407826

2. Start VStrains:

vstrains -a spades -g spades_G_SRR29407826/assembly_graph_after_simplification.gfa \
-p spades_G_SRR29407826/contigs.paths \
-o vstrains_G -fwd SRR29407826_1.fastq -rve SRR29407826_2.fastq

Files with reads:.
SRR29407826.zip
VStrains log:
vstrains.log
Spades log:
spades.log

[RunpengLuo]

Hi, Thanks for trying VStrains and sorry for the late reply,

1. the GFA version conflicts with the external python library for parsing GFA file (gfapy) and it is not up-to-update with GFA version 1.2, I'll try to fix it later on but currently changing the version to 1.0 manually from GFA file should work for parsing. Thanks a lot for pointing this out!

2. VStrains didn't test with coronaSPAdes but mainly on SPAdes. I've ran your dataset with SPAdes (common version) + VStrains. It might not be helpful to run VStrains with coronaSPAdes since it already report a single collapsed strain and the graph structure doesn't have edges to further process either. I've attached the result&Bandage visualization if it might be helpful.

Feel free to let me know if there are further questions,
John

out_SRR29407826.zip

[Dv1t]

I've also tried --rnaviral option of spades and there are two types of outcome with VStrains in case kmer=0:

1. Strain assembled (but same as scaffold of spades)
2. VStrain runs forever

Attaching logs and files for both cases:
First:
vstrains_kmer_0_success.log
spades_kmer_0_success.log
kmer_0_success_reads.zip

Second:
vstrains_kmer_0_fail.log
spades_kmer_0_fail.log
kmer_0_fail_reads.zip

[RunpengLuo]

I think it is likely due to the input graph is already disconnected since inferred k-mer size=0 and only 3 rows in the GFA file, and VStrains does not handle disconnected graphs (no edges) since no further graph simplification can be done post SPAdes --rnaviral. I've pushed the changes to make sure VStrains exit when such case exists. The key reason for running forever is due to paired-link inference attempts to infer (k+1)-mer linkage from paired-end reads, when k=0, inference step infers all 1-mer linkage, such search space is infeasible and won't provide any useful information.