End to end alignment and Local in Bowtie Options

[www1151]

Hi there,

Is there a reason that HiC-Pro is run in end to end alignment mode over local? Would changing the default settings to local not be recommended? 

Best

[Jiang Xu]

Hi there,

    My understanding is that HiC-Pro runs  global first then local. So what is the question?

Best!

Jiang

--
You received this message because you are subscribed to the Google Groups "HiC-Pro" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hic-pro+u...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/hic-pro/61459e15-612b-47c9-9fb6-94eb409aee64n%40googlegroups.com.

[www1151]

Hi there,

Perhaps my choice of wording was a bit poor, but Bowtie 2 is run in end to end mode for both step 1 and step 2 of the alignment procedure. When I refer to local I mean the local alignment mode of Bowtie 2, and not the "global" and "local" options in the HiC-Pro options. As I understand, step 1 of the mapping (global) refers to standard alignment of the reads, and step 2 (local) is an attempt to rescue reads crossing the ligation junction. End to end mapping is still used after trimming the portions of reads crossing a ligation junction in both cases as I understand. 

So the question is whether or not it would not be best practice to run bowtie 2 in local alignment mode over end to end. 

Best

[nservant]

Hi all,

To make a long story short, the first version of HiC-Pro indeed run first a global end-to-end mapping, and then a local mapping to rescue the reads spanning ligation junctions.

However, we then realized that the local mapping is not the best strategy (at least with bowtie2) because you cannot force the alignment to start at the 5' end of the read ...

So we decided to change a bit the strategy with ;

- A global end-to-end mapping in step1

- Then, for the reads which do not align, HiC-Pro looks for the ligation motif and trims the 3' end of the read. Therefore we are sure to keep the 5' end of the reads for the step 2.

- And finally, step2, we then redo a mapping (global end-to-end) with the trimmed reads.

That's why finally, even if the step2 is usually called 'local' (mainly for historical reason), in practice, this is still a global alignment on the trimmed reads.

Does it answer your question ?

All the best

Nicolas

[www1151]

Hi Nick,

Thanks for the reply. My question was on whether or not it would be recommended to run step 1 and 2 of the mapping with the --very-sensitive-local option in bowtie 2 over the default options for end-to-end mapping in order to increase alignment rates. For some reason, using the global alignment settings, only my second read pair fails to align efficiently, and the resulting output is less than desirable. I found that both of my read pairs align efficiently when I use a local alignment through the --very-sensitive-local flag. Are there any reasons not to do this?

Best

[nservant]

Hi,

Did you check if you could have any adapters issues on R2 ?

Regarding the bowtie2 options, you can change any options, but I would definitely advice to keep the end-to-end mapping.

In local mode, bowtie2 will report the longest local alignments, which does not necessarily starts at 5' end of the read. This could have an impact on impact on the contact maps (although they should not be that high).

I'll try to give you an example with a ligation product (between two fragments A and B)

                        R1                            R2

        ***************>         <************              

       |--------|---------------------------------------------|

            A                            B              

In this case, the interaction should be between A and B.

With bowtie2 in local mode, it is likely that your R1 read will be align on B (and not on A), just because the part of the reads which aligns on B is longer than the one which aligns on A.

In this case,  HiC-Pro will classify the interaction as dangling-end

That's why HiC-Pro trimmed after the ligation junction and then use a end-end mapping.

                        R1                            R2

        ****>                           <************              

       |--------|---------------------------------------------|

            A                            B              

Hope it helps

Nicolas

[www1151]

Hi Nick,

That helped! I was suspecting that running the alignment with that local flag might be problematic if the reads actually spanned the ligation junction, and I forced it to align by allowing for soft clipping. Thanks for confirming. 

And yes, I obtained the adapter sequences, and the results did not change much after trimming. I also did check for overrepresented sequences in my files, and did not find anything. I suspect there might be some quality issues with the library and/or the sequencing run. Thanks a lot!

Best