chillier
Senior Member (Voting Rights)
A long time ago I compared the findings of this study to those of Germain, Hanson et al 2018. You can see the results below. Agreement between the two studies is moderate and that's a consistent feature of metabolomic studies. I hope to do more like this.
View attachment 21442
This is cool I have a lot of thoughts about this.
First thing is Hoel 2021 use serum and Germain 2018 use plasma. The metabolomes of serum and plasma differ in a few ways, most relevant here is serum have higher levels of amino acids than plasma. It's not totally clear why this is but it's discussed in this paper and they suggest it could the result of proteolysis in serum or from activated platelets (EDTA/Citrate in plasma prevents this from happening). So it possibly follows that the amino acids between these two papers disagree greatly, despite some large fold changes reported.
Second thing is that in Hoel 2021 the majority of patients and controls are non-fasting. I couldn't see them report in Germain 2018 whether they were fasting or not. I would expect big differences between fasting and non-fasting states. Naviaux's paper and Lipkin/Fiehn's paper for example both have patients in the fasting state so I would not be suprised if there's disagreement between these sets of papers.
Also, are all of these metabolites here reported as significant in both of these papers? If there are metabolites that aren't significant included then you might expect them to be governed by sampling bias/noise and I would not expect these to correlate well between studies necessarily.
Yeah the different names of metabolites are a nightmare. I used a "fuzzy matching" program and then manually double checked to make sure it wasn't mixing up things. Some studies provide CHEMID and other meta-labels for the compounds. But there's more than one standard for labels and not every standard includes every metabollite! I'm not able to be sure i've found every match.
I see you also use R for data analysis! Possibly there's a package in R for matching metabolites. I haven't seen it but I wouldn't be surprised.
There seem to be a few R users here. I saw a few ggplots in the thread on EEfRT. Perhaps we could set up a github for sharing code to run analyses.
Good idea, all for sharing code! Have you tried the package MetaboAnalyst? Might be able to do what you're looking for. It can handle KEGG/CHEMID/ compound names and surely has a function to switch between them.
It makes me happy to see your choice of theme for your ggplots @Murph, and your use of labs and captions![]()
I don't remember naming conventions in these studies, but are they the same so you could compare all the compounds they tested for without having to do some changes? I had to do a workaround with metabolic workbench to standardise metabolite names a few years ago when I wanted to compare findings from metabolomic diet pattern studies.
I wish more studies would provide their data so we could do these types of things ourselves..
I Agree! I am a theme_minimal() appreciator.
Not sure how often this would be possible with the available data, but sub-grouping by sex may be helpful.
I think this is a good idea. In multiple metabolomics papers there seem to be big differences between the sexes. In Hanson lab's longitudinal 2 day CPET study the glutamate phenotype they report is almost exclusively a female one.