Running viralVerify on metaSPAdes gave out much more contigs compared to metaviralSPAdes output

[lijingdi]

Hello developers, I have a bunch of metagenomics samples and I want to assemble and identify both bacterial and viral contigs from them (ideally plasmid also). So I ran SPAdes to assemble a test sample using different modes: "--meta", "--metaviral", "--metaplasmid".
What I found are:

1. Result from using "--meta" mode did not overlap with the results from either --metaviral or --metaplasmid modes, which makes sense to me. Results from "--metaviral" and "--metaplasmid" showed some overlap, which makes sense too as the former output both linear and circular extrachromosomal contigs and the latter output circular extrachromosomal contigs.
2. I then pooled the contigs from all three assembly modes and ran viralVerify (with BLAST against the nt database) to classify them as viral or plasmid. I initially expected that viralVerify would not classify any contigs from the --meta assembly as viral or plasmid. However, I was surprised to see that viralVerify classified over 100 contigs as viral, while the --metaviral assembly had produced only 19 contigs.

Now I'm a bit confused about the results. Do you think those >100 contigs identified by viralVerify but were not assembled by metaviralSPAdes be false positives?
Additionally, is there a recommended way to assemble both chromosomal and extrachromosomal elements in a single run using SPAdes? So that I could try to do the viral identification on the pooled contigs set using different pipelines.
I’d greatly appreciate your thoughts and any best practices you can suggest.
Thank you so much for your time and help!

Best,

Judy

[asl]

@Dmitry-Antipov

[Dmitry-Antipov]

Now I'm a bit confused about the results. Do you think those >100 contigs identified by viralVerify but were not assembled by metaviralSPAdes be false positives?

1. viralVerify should be quite outdated at this moment, so false positives are possible. I'd do the verification with something more novel. This paper https://link.springer.com/article/10.1186/s40168-023-01533-x has some benchmarking of different tools with similar purpose; likely there are more novel ones.

2. metaviral/metaplasmid purpose is an attempt to reconstruct complete extrachromosomal sequences that are identified as specific circular or linear high-covered paths in the assembly graph. Viruses with gaps in coverage or complex internal structure can be reported in fragments by metaSPAdes but completely missed by metaviral. Also it's quite possible that some of metaSPAdes' viral contigs are fragments of one contig in metaViralSPAdes' assembly. So the situation you described should be normal even with the perfect viral identification tool.

Additionally, is there a recommended way to assemble both chromosomal and extrachromosomal elements in a single run using SPAdes? So that I could try to do the viral identification on the pooled contigs set using different pipelines.

Unfortunately no such way is implemented in SPAdes

[asl]

Additionally, is there a recommended way to assemble both chromosomal and extrachromosomal elements in a single run using SPAdes? So that I could try to do the viral identification on the pooled contigs set using different pipelines.

You can start metaviral / metaplasmidSPAdes from the assembly graph created by metaSPAdes. See https://ablab.github.io/spades/running.html#other-input Note that the output will be different.