Reads that don't map back on assembly itself

Question

[Bwremjoe]

Hey all,

I recently made a mistake in a script, and it still confuses me what exactly happened. Since part of my analysis relies on some of these details, perhaps some of you can elaborate.

I recently assembled a metagenome (metaspades) and mapped the reads back onto it (bwa). Usually, not all reads map back, because not all reads are assembled. So far so good.

However, I recently accidentally re-assembled only those unmapped reads, and some of them assembled into 1000+ basepair sequences. Can anybody explain to me exactly how that can occur, since they (apparently) did not assemble the first time? How come they assemble now?

Thanks a lot! :)

Answer 1

[Istvan Albert]

The way of the assemblers are mysterious - we may ask why something happened but usually, the answer will not be forthcoming :-)

One way reads can be left unassembled is if they cannot be resolved within the existing larger and better assembly. if a read fits in many places but everywhere is unreliable

once you separate the reads, now the assembler has to do with what it got and builds a much worse model everywhere

long story short, assembly is a complex and not all that easy to trace protocol

[Bwremjoe]

Ah, thanks for this answer. It's more or less what I expected, but it's really making my life difficult right now :')