minia assembly strange result

[yzzhang]

Dear all,
I am a newbie in de novo assembly. I just used software minia 1.4683 to assembly my data (plant genome about 380 Mb,Hiseq 2000, 101 bp reads, totally 139 million paired end reads). At first, I trimmed the low quality reads and adapter contaminated reads , after trimming, the average reads quality is above 20 revealed by fastqc.
But the assembly result from minia is strange, I tried Kmer 31,35, 39, 43, from the assembly results, most contigs started with ploy A (about twenty bp), and with the Kmer grow bigger,the GC content grows down . Does anyone face things like that?
Thank you!
yun

[rchikhi]

I communicated via private messages with yun.

I checked a human assembly made with Minia, and indeed, the first contigs outputted often start with polyA. This is not an artifact of yun's data. It happens when the polyA kmer is present in the data, and this is NOT an assembly error. This has to do with how Minia it chooses a starting kmer, and should not impact assembly correctness.

To verify this assertion, I just computed full-length contigs alignments (in my case, to the human genome) to make sure that the ployA start of the contigs was correct. Visual inspection of the first dozen of contigs shows that it is.

So, to sum up, that polyA start is normal and the contigs are likely correct.

The slight variation of GC content should not be specific to Minia, and is likely a consequence of lower coverage due to higher kmer length.