Do you think it's possible to combine short read Illumina libraries (WGS) from multiple closely related eukaryotic microbial strains (e.g. libraries from a re-sequencing study, >99% ITS1 sequence) to boost read coverage during a de novo genome assembly run (e.g. with SPAdes)? My target genome is much larger than anticipated, and I'm wondering if increasing read coverage by an order of magnitude could improve length accuracy, contiguity, etc.

Maybe differences in repetitive regions or transposable element movement prevent such an approach? Or maybe kmer frequencies are affected enough that de brujin graph construction gets messed up.

There seem to be several pitfalls, but maybe benefits could outweigh the costs here. If anyone else has done this or considered it, I'd be interested in your thoughts. Thanks!

  • $\begingroup$ Please clarify your specific problem or provide additional details to highlight exactly what you need. As it's currently written, it's hard to tell exactly what you're asking. $\endgroup$
    – Community Bot
    Nov 7, 2022 at 22:33

1 Answer 1


Given that the differences between your strains probably involve rearrangements and copy number variations, this approach seems inadvisable if your goal is a highly contiguous genome. You will make the dBG into more of a hairball if such differences exist.

Illumina is quite cheap these days, I would vastly prefer simply sequencing more deeply if your issue in inadequate coverage. Online tools exist for figuring out how to get there.

Unfortunately, you will swiftly plateau in terms of contiguity even for quite simple genomes using Illumina. So you can expect quickly diminishing returns.

Knowing more stats on your assembly would help evaluate this. For example, if you are working with effectively 10X coverage with N50 ~ 500bp, I might be willing to look the other way while you co-assemble such a synthetic metagenome, because you will at least be able to assemble gene length contigs, which are unlikely to vary too dramatically between your strains. However, if the issue is that your assembly has an N50 ~ 10Kbp and you want it to be higher, then you should consider using other approaches to improve your assemblies, such as long reads. In a pinch, you could do something like optical maps or Hi-C, though I believe that is not preferred these days given much cheaper/better long reads. (Full disclosure: I used to work at a Hi-C company and still hold stock.)

The field of metagenomics is based on the assumption that your suggested approach is ok to get a basic idea of e.g. gene complement, and if you are clever you can even deconvolute such mixtures into genomes, under the assumption that different genomes are at most distantly related. But that doesn't mean that it's a good idea to do it in silico by combining isolates.

  • $\begingroup$ Thanks for your thoughts. I actually do have PacBio HiFi contigs as well, but hybrid assembly efforts have only gotten me so far (tho much higher N50 than you mention, which is nice). Also not in position to continue sequencing :( I'm sorry if this is too speculative for this forum, I've just never seen anyone publicly entertain this notion. I think i agree that broader rearrangement patterns will likely hurt more than the process might help. $\endgroup$
    – bishopia
    Nov 8, 2022 at 13:40
  • $\begingroup$ @bishopia I would say that if you have HiFi contigs, go with those! You are not going to improve them by adding a bunch of heterogeneous shotgun Illumina data, as you seem to have inferred. $\endgroup$ Nov 9, 2022 at 17:56

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.