'N-Blocks' or Any Nucleotide Alignments Best Practices

[wilflugo]

How 'N-blocks' in a genome are normally handled on a Local and/or global alignment computation?
I have been reading about this but seems I can't find what the 'best' practice should be. I see some tools that remove all N-blocks from a chromosome before doing the alignments and others assign a penalty whenever a nucleotide is compared against a 'N' nucleotide.

As far as I understand (and I may be wrong) 'N' is an unknown nucleotide that has not being decoded. Thus in theory could be any one. However, if I assigned a match on a large N-block region, all individual reads are mapped incorrectly(?) into that region since the alignment see an exact match with all the highest scores into that region.

Should N-blocks be simply trimmed out of the chromosome? Is there any other methodology to follow when doing alignments between reads and chromosomes (Assembly)?

Sorry if this is common knowledge, but turns out that searching for 'N-blocks' is not a wise search key since it matches almost everything that is not relevant to this topic.

thanks.

[Brian Bushnell]

Aligners generally ignore large blocks of Ns, or replace them with random sequence (which is not expected to match a read). When you align reads to chromosomes, just leave the Ns there, they won't hurt anything. If you remove them, your coordinates will get messed up.

[wilflugo]

Thanks!. Haven't thought about the random sequence. Thank makes sense and it is probably the way I will be going.

However, I am a little bit curious, how can you ignore N blocks when using optimal local alignment algorithms like Smith-Waterman where scores are based on match/mismatches on a pair of nucleotides?

You could score 'zero' on all Ns, but depending on the similarity matrix used, zero could be a valid penalty value. I guess that ignoring a block need to be addressed case by case and depending on the statistical significance of the similarity matrix.

thanks again.

[Brian Bushnell]

BBMap, for example, assigns a positive score for a match, negative score for a mismatch, and 0 if either the reference or read are N. For nucleotide alignment similarity matrices are not used; those are for amino acid alignment.