de novo metagenomics on viral NGS data is a hot-topic. On this site alone at least 4 specific algorithms have been used to identify multi-strain/multi-species for a given data set, however these do not necessarily involve short-read data, and are listed below.

Questions What approach is recommended for traditional Illumina multiple infection data both for:

  1. for intraspecific (intra-species) data; and
  2. for different species co-infecting the patient, but within the same virus family?


  • Benchmarking using a common standard (for example)
  • Error rates, not miscalling a SNP but producing mosaic artefacts (an artificially mixed genome)
  • Suitability of short-read data for intra-species metagenomics.

Programs discussed on Bioinformatics SE:

SAVAGE https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5411778/ Baaijens et al. De novo assembly of viral quasispecies using overlap graphs Genome Research 2017 May; 27: 835–848.

S-aligner (@juanjo75es) https://www.biorxiv.org/content/biorxiv/early/2021/02/02/2021.02.02.429443.full.pdf Juanjo Bermúdez s-aligner: a greedy algorithm for non-greedy de novo genome assembly 2021 bioRχiv February: 1-7

Rey http://denovoassembler.sourceforge.net Sébastien Boisvert et al Genome Biology. 2012 13:R122, December

@LRJoshi here used the Broad Viral NGS pipeline for de novo assembly. https://viral-ngs.readthedocs.io/en/latest/assembly.html

These are very small genomes < 30 kb. Quasi-species is a very real phenomenon (loads of mutations), however mixed infections are also well-known. In a mixed infection, producing one genome which is a mix of two genetically distinct viruses is a key concern.


1 Answer 1


Under my experience, the best ones in completeness and N50 are s-aligner, Megahit and rnaSPAdes, but I admit I didn't test the other ones for different reasons. Broad indeed seems to use internally SPAdes and Trinity. I compared the results from these three with reported results in benchmark sets from the other assemblers and the results were better for these three.

From these three I get the best results with s-aligner but it has some issues: it actually requires some practice using it and it is slightly slower for this kind of data (<30kbp genomes). In the other hand, it usually has lower requirements in memory than other assemblers except maybe Megahit.

Here there are some comparative results (NGA50) using public benchmark sets: Small virus assembly benchmark Large virus assembly benchmark

Obviously, I am kinda biased as I am the author of s-aligner but the data doesn't lie.

Getting the best results also may not be the first requirement in some cases, like if you want it fast and with a fast learning curve, and in this case Megahit I think is usually quite good. But in these cases, you also can design a script that will automate s-aligner for your specific needs. You can contact me if you need help.

BTW, s-aligner also gets better results with other kinds of data like bacterial genomes, transcriptomes, or small eukaryotes but always at the cost of requiring more running time or more computational resources (running it in AWS, for example).

  • $\begingroup$ Many thanks @juanjo75es. The key take home message thus far is the consistency of the benchmarks used and s-align does resonate strongly on this index. I'm considering this from a conservative point of view. The variable performance of Spade/MetaSpades is weird (everyone is aware of Spades). Run time is not the issue as you've pointed out. I've some reading to do. $\endgroup$
    – M__
    Mar 9, 2022 at 21:07
  • $\begingroup$ Great. Indeed with the last version t actually gets better results than what it's reported in the document and does not require Flye even for larger viruses. If you want to test it at some point, let me know and I will provide some advice and the last version. $\endgroup$
    – juanjo75es
    Mar 10, 2022 at 2:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.