How to evaluate assembly quality?

[sunyu1357]

Hello guys,

After I played around with Mosaik aligner for a while, I started to wonder a question: how to evaluate the quality of assembly after increase mismatch number. Since mosaik does not have a limitation for mismatches in the reads mapping to the reference, I found that if I increases the mismatch, my consensus always get longer, and number of reads mapped always get increased (which also make sense). But how do I know it is not achieved the improvement by jeopardise the assembly quality, such as mapping a lot of reads into wrong position. Is there any way to evaluate the assembly quality?

Thanks!

[sunyu1357]

I can give a little bit detail here: first I assembled 55 paired end solexa reads into reference with 3 mismatches, then 6 mismatches, then 10, then 15, the consensus get longer and longer. But I worried about this comes with a trade off: that a lot of reads mapping with wrong positions.

The distance between sequenced genome and reference genome is kind of medium, ds value for most genes are between 0.08 - 0.13.

[francesco.vezzi]

Every time you start from the same reference sequence or you iteratevely reuse the result of the previous step? I mean do you first align with three errors and then you use the consensus that you obtain as a new reference and youallow 6 errors on this?

I think that is not a matter of tradeoff, but of what are you aligning. If you are alinging against your reference reads that belong to a nhigh releted organism then you have to allow 3 o r 4 mismatches, while if the reads belong to a distant organisms you have to allow more mismatches.
I think that the iterative approach could help you

Francesco