It is possible to improve my assembly?

Question

[emile.cavalet_giorsa]

Hi,

I'm working on the assembly of a plant genome. I have 10X genomics reads (coverage: 60x) and Illumina PE (coverage: 60x).

I've tried to do the assembly with Supernova (10X genomics software and it worked fine). I've got 47000 scaffolds and N50 around 50000.

I would like to improve my results and I have assembled two addictional genomes using soapdenovo2: the first using only the Illumina PE sequences and another using the Illumina PE sequences and the 10X reads without the linker. I've made those two assembly in order to try to scaffold them with the scaffolders that use 10X linkers. I've tried ARKS, ARCS and Scaff10x, but no improvement occurs. I've also tried to raise the N50 merging the assemblies (with CISA and GARM) and scaffolding after that: not a single improvement.

Now I ran out of ideas and I would like to know if somebody knows how I can improve my assembly.

Thank you in advance
Emile

[h.mon]

I don't have experience with 10X genomics technology, but I think it is unlikely you will get a significant improvement over the Supernova assembly using the very same data and ignoring the linked reads.

To improve your assembly, I think you need a good, high density genetic map, and / or additional PacBio or NanoPore sequencing.

[Carambakaracho]

I can't add much to what hmon commented besides some own experience :

I tried years ago something similar and failed. What convinced me at that time was aligning the assemblies and I wasted quite some days until I reached the conclusion.

At that time I learned that the N50 alone is a poor quality metric. I recommend the assemblathon paper and Quast, in case you haven't read them yet. It won't make your assembly any better but maybe it helps to accept it as good enough

Answer 1

[rando2]

Are you working in a species with closely related species that have finished or high-quality assemblies? I ask because there are some tools for reference-based assembly of larger fragments (e.g., chromosome fragments) from scaffolds. The one I have worked with, and had pretty good success with, is Reference-Assisted Chromosome Assembly. The code is here. It doesn't require a genetic map, just a phylogeny.