search

NCATSTranslator / Feedback

A repo for tracking gaps in Translator data and finding ways to fill them.
7 stars 0 forks source link

Arginine is not a treatment for hypercholesterolemia (Translator overweights old evidence and ignores new evidence) #891

Open jaredroach opened 3 months ago

jaredroach commented 3 months ago

I tried looking at results for drugs that treat common conditions - where there are several classes of drugs that treat this condition and many dozens of drugs on the market or in testing. The results for "migraine disorder" looked really good. Pretty much all the classes represented among the top few pages of results, including both modern and classical drugs. I didn't see any false positives. But for "hypercholesterolemia" - not as great.

https://ui.test.transltr.io/main/results?l=Hypercholesterolemia&i=HP:0003124&t=0&r=0&q=a10e49c6-9862-4eb4-ae08-d006421a510a

For example, "arginine" is #8 in the results. I don't think arginine treats hypercholesterolemia. It has been bandied about in the literature, but it definitely does not belong in the top 8.

There are a couple of papers that mention it. Notably an opinion paper from 2007 that sugegst arginine is a substrate for NO in the International Journal Of Cardiology, (Elsevier), so not complete junk. The other 8 papers offered as evidence seem to be off base. E.g., "Effect of pravastatin on responsiveness to N-monomethyl-L-arginine in patients with hypercholesterolaemia". A key paper not mentioned is a 2019 systematic review (PMID: 31922465) that concludes: "A systematic review and meta-analysis found that L-arginine supplementation did not significantly affect LDL-cholesterol, total cholesterol (TC), or HDL-cholesterol levels. It did, however, show a significant reduction in triglyceride (TG) level"

So it is OK to have arginine (pretty much based on a single (later disproved) 'speculation' opinion paper) but it needs to be near the bottom of the list. There are hundreds of hypercholesterolemia drugs that merit higher ranking than arginine.

Maybe underlying problems are:

  1. Over reliance on SemMed DB for ranking results
  2. All but one of the 9 publications used for evidence are from 2007 or before. The most recent is from 2017. So a "fad" idea that was once hot but later disproved somehow (A) gets ranked high, and (B) the counter evidence is not presented. PMID: 31922465 should really show up as part of the "evidence" to balance out the "yes it works" papers.

Screenshot 2024-07-30 at 12 35 20 PM

Note: Issue created in pursuit of this task for teams Expander Agent & Multiomics Provider, as mentioned 7/19/24 in the TAQA meeting notes https://docs.google.com/document/d/1eWwa-sC5xxghu7SNPsXEmUvis52njpT6-2YNzjIr-vA/edit: In addition to those universal priorities, we are asking each team/priority owner to take ownership of their work throughout the release process. That includes: Manually testing in TEST https://ui.test.transltr.io/.and creating issues as needed in the Feedback Repo

sstemann commented 3 months ago

do you know if this is scheduled for a release to CI - Guppy or Hammerhead maybe?