why consed finds discrepancies with masked n sequence

[alig]

Hi,

I have just received our first Illumina data & I have assembled it into Consed. Then I searched for high quality discrepancies as am looking for variants.

But also in the list brought up it has discrepancies with 'n' (masked) sequence in the RefSeq.

Why would the program compare the sequences to regions of N's . I would have thought it would only be looking for differences between the 4 bases.

How can I screen these out to reduce my long list of discrepancies?

Thanks alig

[basickler]

If you're using Phrap/Consed you might want to think about using something like MAQ or similar programs as Phrep/Consed will not do well with short read sequencing (doesn't really do well on anything but sanger BAC re-sequencing/assembly without tweaking it a lot really). If you're working on de-novo stuff and don't have a reference try velvet or euler-sr.
Cheers,

Brad

[sklages]

Consed uses cross_match to align the reads (454,illumina) against a reference sequence,
phrap is not invoked. If you are "working" with the alignment, e.g. some kind of finishing, then
you are somewhat dependent on programs like consed.

But I agree that consed / cross_match is not doing that well on that, mainly because of the
problem mentioned and as well as it (still) lacks some proper consensus recalculation after
(short) reads have been aligned.

So if you're just looking for discrepancies, I'd also recommend programs like MAQ or
Mosaik Assembler (from the Marth Lab) which is well suited for SNP searches as well.

hth,
Sven

[bioinfosm]

cross match seems pretty sensitive to aligning short reads correctly. It is a matter of obtaining SNP calls from it and then compare.

The discrepancies reported by consed need to go through some sort of filtering to be able to get good quality snp calls.. anyone worked with this?

perhaps cross match mapped reads -> maq map format -> maq SNP calling?