Distal species alignment | chainCleaner & chainPreNet

[RocesV]

Hi everyone! I am a noob in these Comparative Genomics stuff so excuse me if this is a obvious question.

I am working with two distal species (181 MYA) and some of your tools looks very interesting. I have checked "Supplementary Material for a genome alignment of 120 mammals highlights ultraconserved element variability and placenta associated enhancers" and it seems like you did not use chainPrenet before netting (chainNet) ,after chainCleaner.

• Are these two tools mutually exclusive?
• Is a bad idea chain-filtering (chainPreNet) after chain-improvement (chainCleaner) to obtain better nets?
• Should i use always chainPreNet before chainNet?

Thank you in advance! 👍

[MichaelHiller]

We run chainPreNet after RepeatFiller and after chainCleaner (which are the two steps to improve the chains), but before running chainNet.
My understanding is that chainPreNet filters out chains that would be 'netted away' anyway, making chainNet to run faster and likely with less memory. I have never tried to leave it out and compare the differences.

Hope that helps.
Michael

[RocesV]

Thanks for the quick response! I got it. One last question:

Does it make sense to combine patchChain and RepeatFiller in the same workflow?. Both of these tools are based in more sensitive alignments but doing some comparisons i could appreciate that RepeatFiller dont add more Mb to my .chain when i used patchChain before.

Using patchChain + RepeatFiller -- + 10.9 Mb to .chains file
Using patchChain -- + 10.9 Mb to .chain file
Using RepeatFiller -- + 0.2 Mb to .chain file

😄 !

[MichaelHiller]

Good question. We have never tested that.
The difference is that patchChain computes new local alignments keeping the masking and then generates chains from old + new local alis. This can sometimes change the chains a bit.
RepeatFiller removes the masking (but I think we have a flag to disable that) and strictly adds the new alignments to the existing chain. An issue is that the longer the unaligning region, where you run another round of highly-sensitive alignment, the more likely it is that you get alignments between non-orthologous repeats. This is why we used larger thresholds for how long the unaligning region can be for the patchChain workflow and smaller default thresholds for RepeatFiller.

But give it a try and have a look if RepeatFiller after patchChain adds meaningful (likely orthologous) alignments.

[cfz1998]

Hi! @MichaelHiller
I was mapped the apple genome to pear genome. When i running

chainPreNet all.chain.gz reference/t.sizes q.sizes all.pre.chain

I got a empyt result file: all.pre.chain.
It likes innormal! Why?
Any advise will be helpful. Thank you for your reply!
Best wishes!
--zcf

[MichaelHiller]

Hmm, we never had such a case. I guess t/q.sizes are not empty?
Definitely this is not normal. ChainPreNet typically removes only very small chains that overlap longer chains.

Pls check if your all.chains look fine and contain long syntenic ones.