Using SPAdes to assemble genomes from historical DNA

[LiaOb21]

Dear SPAdes developers,

I am currently working with genomes from historical fungal samples and using SPAdes, along with other tools, for assembly.

I have a few questions about how best to run SPAdes in this context. As you can imagine, the DNA is degraded, fragmented, and coverage is uneven.

I have tried assembling the genomes using both paired-end Illumina reads and merged reads (in separate runs). Based on BUSCO scores, the differences are minor, and neither input type consistently produces better results. Would you recommend using paired-end reads rather than merged reads, or does this not make much difference for the algorithm?

I am also curious about how SPAdes handles reads shorter than the smallest k-mer in the list. Are these reads discarded? Do you have any recommendations for selecting k-mers? So far, I have tested both the "auto" mode (21, 33, 55) and a longer list (21, 27, 29, 31, 33, 35, 37, 41, 51, 61, 71, 81, 91, 101, 111, 121). The BUSCO results are very similar in most cases, so I wonder whether including a more extensive set of k-mers provides any practical benefit.

Finally, regarding the --isolate flag: the log suggests using it, but I have been hesitant given the probability of contamination in these historical samples. Do you have any recommendations on whether to enable this flag in such cases?

Thank you very much for your time and advice.

Best wishes,

Lia

[asl]

Based on BUSCO scores, the differences are minor, and neither input type consistently produces better results.

I do not think BUSCO is a good metrics of assembly quality frankly speaking.

Would you recommend using paired-end reads rather than merged reads, or does this not make much difference for the algorithm?

In general we do not suggest using merged reads as things could be mismerged, etc. You do not want to fight with merger bugs / problems in addition to anything else.

I am also curious about how SPAdes handles reads shorter than the smallest k-mer in the list.

These are simply discarded as there is no way SPAdes could extract any k-mers from these.

Finally, regarding the --isolate flag: the log suggests using it, but I have been hesitant given the probability of contamination in these historical samples.

Well, have you checked that you indeed have this contamination? And it is highly abundant and could interfere with results?

[LiaOb21]

Hi @asl,

Thank you for your reply.

Given the level of degradation of the samples, we are taking into account BUSCO scores primarily, as we aim to recover as many ortholog genes as possible, but we are also checking assembly metrics for contiguity.

We do have contaminants in the samples, and generally, they have higher coverage than the target genome.

Do you have any recommendations on the k-mer list? Do you think it is a good idea to use as many k-mer sizes as possible or not really?

Additionally, do you recommend trying any other parameter for this specific use case?

Thank you so much again.

Lia