dataset: DISEASES#66
Conversation
tristan-f-r
added
dataset
Co-authored-by: Neha Talluri <78840540+ntalluri@users.noreply.github.com>
this is only in github actions
Co-authored-by: Neha Talluri <78840540+ntalluri@users.noreply.github.com>
to move to web
…ed numbers in the README.
|
The committed change now includes using the full DISEASES datasets. We now have 121 disease datasets, though we could filter this (either by increasing the minimum number of high-confidence gold standard genes in the disease set, or by requiring a certain number of prize nodes). |
Does that mean that the diseases code should be trimming, or does that happen elsewhere? That might change the new numbers, if we require at least 10 gold standard disease-gene pairs that appear in STRING. |
| There are 84972 high-confidence disease-gene pairs from the 643 diseases | ||
| ``` | ||
|
|
||
| (Note: if you use the filtered datasets, you end with 134 diseases with at least 10 high-confidence disease-gene pairs). |
There was a problem hiding this comment.
The snakemake file only mentions 121 diseases
There was a problem hiding this comment.
There are 134 diseases that have at least 10 high-confidence disease-gene pairs from DISEASES. Some of these don't have enough TIGA evidence to be used - that's in the next step. At this point, there are 134 diseases that COULD be used.
| diseases_path = Path(dir_path, "..") | ||
| (diseases_path / "processed").mkdir(exist_ok=True, parents=True) | ||
|
|
||
| OXO_URL = "https://www.ebi.ac.uk/spot/oxo/api/search?size=500" |
There was a problem hiding this comment.
This site is down right now and so I can't run the code to process the diseases data. Is there a way to download this data and have a static version we use?
There was a problem hiding this comment.
I put in a help ticket and will wait to hear back. Once it's back up, we should retain a static version.
If it doesn't come back online, I will look up the original data sources (MONDO and EFO) and get cross-reference files from those databases.
…hmarking into diseases-dataset Merging neha's commits after running locally, before updating the OXO parsing code to rely on flat files rather than API calls.
|
Using the OXO2 mapping files rather than the OXO API (which has been taken down) gave us three more diseases, bringing the total datasets in this collection to 124. However, the number of prize nodes and the number of gold standard nodes vary drastically, and I noticed that the largest disease is simply "Disease", the next largest is "Disease of the Nervous System", etc. I will explore the relationship between gold standard nodes and prize nodes to see if we should trim further. As a reminder, we currently identify the diseases in the DISEASES database for which there are at least 10 high confidence disease-gene pairs (scoring a 4 or 5 out of 5 in the DISEASES database). Of those diseases, we retain any that have corresponding TIGA trait-gene scores (based on the SNP scores only; the experimental track is ignored in the DISEASES database to avoid circular reasoning). |
3 participants
From #39.