They did this database in collaboration with Deepmind so they probably agreed on the coloring and the interpretation of the scores. Deepmind people also use this coloring scheme in their AlphaFold 3 server and tutorials: https://www.youtube.com/watch?v=UqeQfRDA8Yk <https://www.youtube.com/watch?v=UqeQfRDA8Yk> (e.g. min. 9) So I would expect that people think about AF pLDDT scores in these colors, where e.g. light blue(or cyan) means above 70. It is just a matter of consistency. Of course if you read carefully the alphafold palette in ChimeraX you can see that yellow is for 70, so you can expect -60-80 to be yellowish. If ChimeraX had come up with a completely different palette people would read and learn what each color means, but if it was meant to look like alphafold coloring scheme, but it is not exactly like it, it is misleading, at least to me. Cheers, Agnieszka
On 15. May 2024, at 20:42, Tom Goddard <goddard@sonic.net> wrote:
Hi Agnieszka,
I don't think the AlphaFold pLDDT coloring is used at all in the original AlphaFold paper. I'm not sure if the EBI AlphaFold Database made it up. Here is an EBI AlphaFold database web page describing it
https://www.ebi.ac.uk/training/online/courses/alphafold/inputs-and-outputs/e... <https://www.ebi.ac.uk/training/online/courses/alphafold/inputs-and-outputs/e...>
Their ranges are blue "Very high (pLDDT > 90", cyan "High (90 > pLDDT > 70)", yellow orange "Low (70 > pLDDT > 50)", orange red "Very low (pLDDT < 50)".
<AFDB_plddt_colors.png> This is somewhat coarse as it goes directly from "high" confidence above 70 to "low" confidence below 70. It seems they think 70 is medium confidence, that is yellow in ChimeraX. The 80 value which you might call medium-high is a greenish color in ChimeraX, not yellow. Here is the ChimeraX pLDDT colors as rendered by ChimeraX <chimerax_plddt_colors.png>
So ChimeraX gives a greenish color for the medium high 80 value and yellow ("caution") for the 70 value.
Of course you can use the EBI AlphaFold database coloring that makes 70-90 all light blue and show their scale that names it "High confidence", even though <70 is also considered "Low confidence" on the same scale. It all depends on what you want to convey in your publication figure. If you want to be optimistic and suggest the AlphaFold model is more confident, then make pLDDT in the 70s light blue. But in ChimeraX I decided to add more shades so you can really see that 70 is medium (yellow) and 80 is medium high (green), and 90-100 is high (light blue to blue).
Tom
On May 15, 2024, at 11:02 AM, Agnieszka Obarska <Agnieszka.Obarska-Kosinska@biophys.mpg.de> wrote:
Hi Tom,
Thanks! One could argue what is more informative but since it is called alphafold palette we blindly expected it to agree with AF colouring scheme. So we thought that yellow means LOW score (below 70) while it is not the case, 80 (HIGH score) can be still yellow. It looks especially weird for AlphaFold3 models, where you have more scores 70-90, rather than 90-100, and regions of HIGH score according to AF, and it is blue on their server, are yellow in ChimeraX. See another example below. We will of course switch to AlphaFold coloring now but I just wanted to let you know.
Cheers, Agnieszka
<PastedGraphic-4.tiff>
On 15. May 2024, at 18:52, Tom Goddard <goddard@sonic.net <mailto:goddard@sonic.net>> wrote:
Hi Agnieszka,
Indeed ChimeraX AlphaFold plddt coloring is a bit different from the AlphaFold database color scheme. For one ChimeraX uses 5 colors while AFDB uses 4 different colors. I agree to best match the overall appearance of AFDB the ChimeraX color scheme should probably put light blue at 80 instead of 90. I'm not sure why AFDB decided to apply single colors to large range of plddt limiting the nuances shown -- maybe it was easier to render in their web pages. At any rate it seems like a poor choice. So I'm not inclined to blindly try to match their color scheme (I made our ChimeraX color scheme). I think the ChimeraX color ranges give a much better indication of the scores including placing light blue at 90 and yellow at 70 so that you get good discrimination of values in the medium confidence region of 70-90.
Thanks for pointing out the slightly more yellow appearance of the ChimeraX plddt coloring. As you showed in your images you know how to make the less useful AFDB color scheme in ChimeraX if that is what you would like to include in your publications. I agree it is unfortunate that a single color scheme was not established at the start that best showed the plddt values. For publications I think a color key would be needed in any case to document what the colors mean rather than relying on uncertain conventions.
Tom
On May 15, 2024, at 9:21 AM, Agnieszka Obarska via ChimeraX-users <chimerax-users@cgl.ucsf.edu <mailto:chimerax-users@cgl.ucsf.edu>> wrote:
Hi Elaine,
Yes, I agree that color by gradient is more informative but I think the gradient you have is not consistent with AlphaFold coloring. When you think about AF coloring, yellow means below 70, so LOW according to AF, while in ChimeraX you can have 80 still yellowish, while it should be already light blue. I think you would need to adjust the boundaries, such that at 70 it becomes blueish. See image attached.
Best, Agnieszka
<PastedGraphic-2.tiff>
On 15. May 2024, at 17:51, Elaine Meng <meng@cgl.ucsf.edu <mailto:meng@cgl.ucsf.edu>> wrote:
Hi Agnieska, The coloring is the same at the boundaries (90,70, ...) but the difference is that the ChimeraX coloring is continuously shaded based on value whereas the one on the right just gives exactly the same color for the whole range (e.g. 50-70 is all yellow on the right, whereas in ChimeraX it is gradually shaded orange to yellow). The continuous shading gives more information on the exact value so that 49 is similar to 51 instead of a totally different color as on the right. I hope this makes sense, Elaine ----- Elaine C. Meng, Ph.D. UCSF Chimera(X) team Resource for Biocomputing, Visualization, and Informatics Department of Pharmaceutical Chemistry University of California, San Francisco
On May 15, 2024, at 6:15 AM, Agnieszka Obarska via ChimeraX-users <chimerax-users@cgl.ucsf.edu <mailto:chimerax-users@cgl.ucsf.edu>> wrote:
Hi,
I’ve just would like to let you know that we’ve noticed recently that alphafold palette in ChimeraX is not consistent with original AlphaFold coloring. Please see the slide attached. On the left there is a model colored by the “color bfactor pal alpha” command in ChimeraX while on the right there is the same model colored according to AlphaFold coloring scheme.
Best, Agnieszka
<PastedGraphic-1.tiff>
_______________________________________________ ChimeraX-users mailing list -- chimerax-users@cgl.ucsf.edu <mailto:chimerax-users@cgl.ucsf.edu> To unsubscribe send an email to chimerax-users-leave@cgl.ucsf.edu <mailto:chimerax-users-leave@cgl.ucsf.edu> Archives: https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/ <https://mail.cgl.ucsf.edu/mailman/archives/list/chimerax-users@cgl.ucsf.edu/>