read alignment trimming

[seqjordan]

Hi everyone,

I want to ask the meaning of the following sentence?
The methylation calls data were aligned using iterative trimming, without trimming any initial bases before alignment.
How should one do the alignment using bowtie2 or Bismark?
Thank you a lot!

[JackieBadger]

The paper doesn't say?

[seqjordan]

The paper doesn't say?

Hi Jackie,

It mentioned: "Based on our M-bias plots we filtered the first 15 bases of each read as part of summarizing the methylation evidence", which means after alignment the first 15 bases are trimmed ? but how does iterative trimming work?
Thanks!!

[simonandrews]

Maybe it's just that they did an untrimmed (or only adapter trimmed) alignment using bismark first. They then saw that the M-bias plot produced by bismark showed artefacts at the start of the reads so they then went back and removed some more bases and aligned again to get a more solid set of methylation calls?

[seqjordan]

Maybe it's just that they did an untrimmed (or only adapter trimmed) alignment using bismark first. They then saw that the M-bias plot produced by bismark showed artefacts at the start of the reads so they then went back and removed some more bases and aligned again to get a more solid set of methylation calls?

@simon, you mean after trimming the first 15 bases then do the alignment again? The methylation call are stored into tsv files and then converted back? What I used for alignment is special bowtie2, as described here: https://github.com/BenLangmead/bsmoo..._align.pl#L207
There is no trimming option in this step. Thank you!

[simonandrews]

We've certainly trimmed and remapped data after seeing an initial m-bias plot to remove odd looking bases at the start of the reads. We used a small custom script to do this rather than a generic trimmer or trying to do it in the mapper. We could just have ignored the first 15 bases when we extracted the methylation calls, but it was easier to trim and remap, and this would also potentially give us better mapping efficiency if the bias was through the introduction of non-native sequence.

[seqjordan]

We've certainly trimmed and remapped data after seeing an initial m-bias plot to remove odd looking bases at the start of the reads. We used a small custom script to do this rather than a generic trimmer or trying to do it in the mapper. We could just have ignored the first 15 bases when we extracted the methylation calls, but it was easier to trim and remap, and this would also potentially give us better mapping efficiency if the bias was through the introduction of non-native sequence.

If bismark is used then the resulting mean coverage is very high, but when I used bowtie2+bsmooth, the coverage is much lower. Why is this so different?

[simonandrews]

Without knowing the details of exactly what you ran it's difficult to say. Both bismark and bsmooth use essentially similar processes to map the data so any differences you see will be in the specific options used - they' both using bowtie (or bowtie2) behind the scenes.