Interspecies alignment and read length comparison?

[BGould]

I am planing on aligning RNASeq reads from a nonmodel plant to a model plant genome. Does anyone have experience with (or know of any great references) comparing the amount of alignment using short (~50bp) reads vs. long reads(~100bp) and the quality of those alignments when aligning to non conspecific reference genomes ?

[chadn737]

I have no experience with this, but I would venture to guess that you would get better alignment using shorter read lengths. The longer the read, the more differences you would have between the read sequence and the reference sequence which would likely result in poorer mapping.

I am curious though, why align to a model plant? I would think that a large percentage of your reads will fail to align. Why not do a de-novo assembly of the transcriptome and compare this to your model plant? Like I said, I have no experience with this because all my work is done on Arabidopsis, which is well annotated.

[lh3]

Longer reads are almost always better for alignment. Nonetheless, I agree that de novo assembly would be preferred.

[chadn737]

Longer reads are almost always better for alignment. Nonetheless, I agree that de novo assembly would be preferred.

Certainly long reads are better, especially if your reads are from the same organism whose genome you are aligning to. They would also be better for de novo assembly. Which I think is probably the preferred method here, but I'm not sure what BGould wants to do with their data.

However, if for some reason, you were wanting to align reads from a different organism to a model organism (without de novo assembly), I think it would be harder to align longer reads because the longer the read, the more sequence difference there will be between the read and the genome you are aligning to. Your aligner would have to allow for more mismatches between the read and reference. Otherwise, a large number of your reads would not be aligned. As an illustration, say you have two orthologs, one from the model organism that you are aligning to and one in the non-model organism, the source of your reads. Say that there is a difference in the sequence every 25 or 50 bps. The longer the read is, the more of these differences there will be in a read. Unless you really played with the settings of a lot of aligners, I would think that this would result in a large number of reads failing to align.

Somebody correct me if my reasoning is wrong. I find this an interesting scenario and would be curious what the best approach is.

[lh3]

Nothing prevents you from doing local alignment.