Explanations Needed with Quantifications about keggcharter HOT 6 CLOSED

Thank you for exposing these issues in such a structured way. It really helps with development, having users report on these issues. And there are indeed a lot of them. Since I am the sole developer, tester and maintainer of KEGGCharter, there are many problems that have slipped by me. I hope more users can report on them as you have, so KEGGCharter can be improved. To answer the problems one by one:

1. I hadn't seen this before. I'll test with the example you provided, and report on here. If it's reproduced, it will be fixed next version.
2. No minimum cutoff is present in KEGGCharter, although it isn't a bad idea. This stems from the bug reported in 1), KEGGCharter filters the input dataset to remove lines with no IDs, here it's considering the absence of value in the Function column as evidence that the row has no information to offer. Once 1) is fixed, this will not be important.
3. Since version 1.0, KEGGCharter has the parameter --distribute-quantification, to divide quantification values by the different K numbers. The division of quantification was the only option before, but now it allows to not do it, and is the default behavior I left it as default as it's easier to interpret directly from the input datasets. Maybe the default behavior should be dividing, nevertheless, the behavior depends on the parameter.
4. This logic makes sense. I'll look into, at minimum, put a more appropriate colormap for those maps. But will try to set it as an option, if I see many different options are available and appropriate, with the default always being a different colour scheme than the one currently implemented, as it is indeed not the best option.

I'm going to work on this either this week or the next. The next KEGGCharter will have this sorted out. The comment about the -qcols and -gcols is also correct, and I'll check in the documentation where it's still not updated.

from keggcharter.

I have fixed everything, except for the quantification values, which KEGGCharter is reporting a bit differently from the result you showed.

Still trying to figure out why. Should be something with the mapping of IDs, since the quantification values are being distributed now.

from keggcharter.

This has been very quick! Sorry I missed the -dq parameter - this seems to be what I want it to do. The new color scale is better - yellow on white is sometimes a bit hard to see, but if it stays that darker shade of it then should be OK. Generally a good idea to avoid shades of green (with close yellow) due to Red-Green colorblindness. Some of the better scales are Red-to-Blue or Yellow-to-Blue. What do you mean above that the quantification values in KC are not working still? The pic you have above seems to be exactly matching the online KEGGMapper version I showed - maxing out at 4000 annd properly showing the 2000 value and then the smaller 20-400 values in darker green. What are you thinking is wrong?

from keggcharter.

About the colormap: I was experimenting with viridis, which seems to be consensually a good colormap, avoiding the green-red colorblindness. However, I felt the darker spectrum of the colors makes it very hard to see the EC numbers/box labels.

Therefore, I decided to keep summer as default, and if people don't mind darker maps, they can now pick viridis or another colormap.

Actually, I would have prefered a lighter map like this but in the colors you mentioned, so as to avoid the colorblindness problem. If you know of such a matplotlib colormap, please let me know. I also tried tinkering with adjusting brightness, spent a couple hours on that, but couldn't make it work for both the map and the side label (only for the map).

About the erroneous quantification values, it was my human eye confusing a three color system with a two color system. Seems good, and thank you again so much for your help in testing KEGGCharter. If you do find another problem, please don't hesitate in contacting again.

I have just released KEGGCharter 1.1, which fixes everything mentioned here. Hope it is to your liking!

from keggcharter.

OK thanks again for your work! In general, I do like the viridis palette and it is quite recommended - it is a shame we can't make the labels in the boxes white, since that would solve the dark box problem (but then be a pain for the yellow ones). I was probably going to to over the labels in a drawing program for the final publication figures, if need be.

I will download the latest version and continue my testing - I'll open a new issue if anything else comes up!

from keggcharter.

PS: I'm trying the new version now (software --help says viridis is default)...one language thing I forgot to mention is that you need to change your references to "input" in the software and wiki to "impute" where you are talking about using a placeholder value (ie: imputation of quantifications or taxonomies). It can be confusing for people especially in the context of software as we are often looking for "inputs" in program packages, and you have input files, but you mean impute missing values in this context.

from keggcharter.