read length impact on alignment

[gene_x]

I recently run into a problem in my ChIP-seq data as it has low alignment rate to the mm9 genome. It turned out that a large number of the reads contain adapter sequences. We haven't applying adapter trimming to our analysis since we can get a decent alignment most of the time. But in this case, I do need to trim the adapters. I have tired cutadapt and it seems to work well.

My question is on how to control the length of the reads after adapter trimming. Does it make a big difference if I don't control anything (i.e. keep everything that is left after trimming adapters?

Thanks.

[westerman]

Shouldn't make much of difference since you are mapping to the known mm9 genome. I'd probably toss everything below 20 bases but that is a personal preference -- I'd rather not think about them mapping repetitively.

[gene_x]

Shouldn't make much of difference since you are mapping to the known mm9 genome. I'd probably toss everything below 20 bases but that is a personal preference -- I'd rather not think about them mapping repetitively.

Yeah, that's what I thought. Is 20 bases based on some estimation?

[hanshart]

Yeah, that's what I thought. Is 20 bases based on some estimation?

20 bases sounds ok for mouse, but even then I would stick to "uniquely" mappable reads, as even for 20bp reads for such a large genome like mm there will be many reads that could be mapped to many valid positions. Have a look at this post http://blog.kokocinski.net/index.php...ability?blog=2 and the therein mentioned papers and online tools for an estimation of a sufficient read length threshold. Of course, you can allow multi-reads to be not so dependent on the (repetitiveness related) structure of mm9 but then your mapping is not as reliable as without multi-reads (IMHO)