looking at some very recent polyploidy events and species splits with ksrates

[sjfleck]

Hello,
Thank you for making this tool. It have been useful for a few research projects already. Does ksrates remove paralogous ks anchor pair data smaller than 0.05? If so, I'm worried it's biasing my results. I have a number of species that may or may not share a recent WGD event so I'm using an x-axis ks limit of 0.25 to look at it closely.

I'm running the analysis multiple times with each species in question as the focal species. Here is mixed_Sa_unadjusted.pdf:

and here is mixed_Sa_anchor_clusters.pdf:

As you can see it looks like the anchor pairs don't have anything smaller than 0.05.

On a separate note, can ksrates do a log transformation? I do a lot of work in CoGe (genomevolution.org) and I'm used to looking at log transformed ks. Any help you can provide is greatly appreciated. Thank you

[Cecilia-Sensalari]

Hi @sjfleck!

Many thanks for reaching out and sorry for the late reply :) Yes, the code removes anchorpoints with a Ks value <= 0.05. This is actually hard-coded and not customizable by the user, at the moment.
See here: function computing the weights of anchor pairs (min_ks = 0.05).
And here: code line that coherently ignores Ks values smaller than 0.05.

I think its meaning was to avoid showing non-biological signal that can appear at ~0Ks. The recent WGMs I had to work with were always located at somewhat higher values, so I never realized it could be a problem! I understand however that this choice may not fit your current dataset.
I'm going to add the option to make this also customizable by the user in the configuration file, then test it and see how it behaves; then release it as v1.1.5.

In these days I'm also finalizing the release of ksrates v2: this latter won't have this latter customizable option yet, but I might as well implement it shortly afterwards. Should you want to try it out, any feedback will be much appreciated!

---

Concerning the log-transformation: ksrates main output files are not log-transformed, but some intermediate files are (especially in the steps for mixture modeling; see the Supplementary PDF for a few more details about it). Do you think that interpreting the log-transformed plot has some advantages?

Cheers,
Cecilia

[sjfleck]

Cecilia,
I understand not realizing how it could be a problem. I've published ksrates plots in Commun. Biol. & bioRxiv (this citation is missing, but it'll be there in the final submission), and there weren't any issues. This is the first time I'm working with very low ks values. I understand the problem with the noise around 0 ks, but if it tests well, I'm looking forward to v1.1.5.

To show another example from a paper that overlaps with a small part of my dataset, this figure also works with very low ks values. I marked where ksrates would filter out their data points too.

I would love to try out v2. I'm working with a separate dataset where low ks shouldn't be an issue, so I would happily test it out.

Regarding log transformed data, I've always felt like it was easier to look at. I can give an example from a recent dataset. On the left is normal ks values and on the right is log transformed ks values. I find the right to be much easier to interpret without needing to change the axes.

Thank you again for developing this program and I'm looking forward to v1.1.5 and v2
-Steve

[Cecilia-Sensalari]

Hi @sjfleck!

I've merged PR #65, which implements customization of the minimum anchor pair Ks value used in the distributions and for the modelling steps (anchor clusters and LMM). Make sure you read the notes about usage I posted in the PR page!

This feature is now available in the pre-release v1.1.5.

I tested it on the example dataset available within the package and it worked out, but I've set it as pre- release because I'd like to hear from you if it meets your needs. Any feedback or comment is welcome!

Also, thanks for the post above on log-transformed data; I will read further into that as soon as I find some time!

Cheers,
Cecilia

[Cecilia-Sensalari]

Hi,
I noticed that the action pushing the updated Docker container to DockerHub failed, and I have to figure out why. I believe it has something to do with building the container for the linux/arm64 platform. Will update this issue!

[sjfleck]

Thank you for the update. I tried downloading it myself, but it still says that it's v1.1.3. I assume this is actually v1.1.4 since that's been the latest version since June 2023.

[Cecilia-Sensalari]

Hi @sjfleck,

Container pushed, v1.1.5! Let me know how it goes.

I tried downloading it myself, but it still says that it's v1.1.3. I assume this is actually v1.1.4 since that's been the latest version since June 2023.

That's possible, I did forget once to update the version number within the _version.py file. However now with v.1.1.5 I did update it!

Cheers,
Cecilia

[Cecilia-Sensalari]

Hi @sjfleck,

Here's another update: I've released v2.0.0, which includes the feature you asked :)

Cheers,
Cecilia

[sjfleck]

Cecilia,
Thank you for helping about a month and a half ago. I ended up needing to focus on other projects, but now I'm getting back to wrapping this up. I have version 1.1.5 and 2.0.0. You said it includes the feature I asked about, but I'm not seeing a min-ks option in the configuration file. Sorry for the delay on this and thank you again for your assistance
-Steve

[Cecilia-Sensalari]

Hi Steve,

Sorry for the delay! In v2.0.0, option min_ks_anchor_pairs was added to the expert configuration file: here you can see how it looks like in the example dataset I provide within ksrates (last line). Is this of any help?

Cheers!
Cecilia

[sjfleck]

Cecilia,
do I add the contents of config_expert.txt on to the end of the regular config.txt or how is it meant to be run?
Thank you,
Steve

[Cecilia-Sensalari]

Hey,

Beside having your "basic" configuration file (e.g. config_myspecies.txt), you can make another "expert" configuration file (config_expert.txt) which contains several parameters to fine-tune your analysis. In this latter, you add the min_ks_anchor_pairs option like shown in the example dataset above!

Afterwards, you launch the commands by taking the expert configuration file as input, as shown in this documentation page.

Let me know how it goes!
Cheers,
Cecilia

[sjfleck]

Cecilia,
thank you for running me through that and I apologize for not looking through the new documentation to see it earlier. I'll let youn know how it goes
-Steve