Why there are some high values in MEME with the plot Beta+ vs Sites?
MauriAndresMU1313 opened this issue · comments
Hi!
I read some articles to interpret the result of MEME; however, I couldn't find out if this is an issue or not.
Well in the plot a little amount of beta+ values are too high, for example, one has a value of 10.000!
Here is a figure of my plot:
This is an issue?
I must delete that value?
I think that maybe is an issue because when a value is too high is a value to ignore.
Anyway, I just looking for some help or recommendation.
Thank!
Dear @MauriAndresMU1313,
This is fairly normal behavior; you effectively have infinite ω rates for some sites. This can occur if you have "saturated" sites, or sites where only non-synonymous substitution occur. Remember that you should never use a point estimate from MEME to assess significance -- always use the p-value reported by the test.
Best,
Sergei
Great to know that! Thank you for your comment! Only in the case that you know some article or protocol to interpret MEME´s result, could you share the name and author here? I read some articles; however, maybe you know the one that I don´t read yet!
Anyway, thank you!
Best!
Hi again, sorry for the insistence. I need to know what to do because I´m plotting my data and I think that for your comment I must take into cont the p-values of my data? I mean plot the p-values, not the dN-dS vs sites? That is correct? Again sorry for the insistence; however, I can´t find out of some article or page to how to interpreting the result of MEME to plot. Here an example of one of my analyses: https://datamonkey.org/meme/60be6abdaeae26084a6f83a3
Thank!
Dear @MauriAndresMU1313,
You can definitely plot dN-dS (or β+/α for MEME). I would suggest you "gate" high values, so for example if this value is >50 you can treat it as infinite.
What I mean by using p-values is that if you wish to say "site X is under selection", you should always consult the p-value for this site, not use β+ values for this determination.
Best,
Sergei
Hi again, I need your help again, well it´s more about your opinion. I plotted some of my analysis in data monkey with 3 different methods: SLAC, FEL, and MEME. Here the plot:
In MEME I plotted the positive selection (beta+), well my question is:
What do you think?
Well in my case I think that there is something weird, I mean there are a lot of values close to 1. This is weird in the sense of natural selection. This a surface protein of some pathogens (viruses) so natural selection act in a strong way in some sites; however, I keep thinking that is weird to have so many sites with values close or equal to 1.
Do you think that my concern is correct?
Any comment is welcome!
Thank! @spond
I forgot: the y-axis are: (1): dN-dS, (2): omega and (3): beta+
Dear @MauriAndresMU1313,
This is not surprising, but is quite counter-intuitive. First, I would plot β+ / α because it is comparable to dN/dS from FEL. You'll have to deal with cases α = 0 by setting the ratio to some large number. Here's why you may find a large number of β+ = 1.
Imagine you have a site which is subject only to negative selection, or is evolving neutrally. The MEME model assumes that the evolution at this site is described by
- Fraction p- at rate β- < α
- Fraction 1-p- at rate β+ ≥ α
If there is no evidence of positive selection at a site, then MEME can do one of two things
- Move β+ to the neutral regime (i.e. β+ = α or dN/dS = 1)
- Set p- = 1, i.e. place 0 weight in the β+ class, which means that β+ cannot be estimated (it will be set to arbitrary value).
I think the sensible way to plot the selected component of the MEME model would be as follows
- If p- = 1, do not plot anything (β+ is not estimable).
- If p- < 1, then plot β+ / α
- Mark sites that have p < 0.05 on all plots.
Best,
Sergei
Thanks for your continue support, I going to follow your advice!