youainti
/
ClinicalTrialsDataProcessing
1
0
Fork 0
Code Issues 3 Pull Requests Packages Projects Releases 6 Wiki Activity

Got counts of competing drugs back

Browse Source
main
Will King 1 year ago
parent 2488cceebc
commit d912408456
2 changed files with 152 additions and 5 deletions
Show Stats Download Patch File Download Diff File

29
Scripts/ConfiguringFormularies.sql
Unescape Escape View File

@ -43,6 +43,9 @@ CREATE TABLE "Formularies".usp_dc_2023 (
"API Concept" varchar(250) NULL "API Concept" varchar(250) NULL
); );
``` ```
It links rxcuis to other rxcuis where they have a matching USP Categories and Class
This gives alternative RXCUIs based on category an class.
*/ */
CREATE MATERIALIZED VIEW "Formularies".rxcui_category_class_links AS CREATE MATERIALIZED VIEW "Formularies".rxcui_category_class_links AS
WITH base AS ( WITH base AS (
@ -72,27 +75,43 @@ I'll' break this into two steps.
1. link formulary alternatives to compounds and brands, 1. link formulary alternatives to compounds and brands,
2. link nct_id to formulary alternatives 2. link nct_id to formulary alternatives
*/ */
drop if exists materialized view "Formularies".match_trial_compound_to_alternate_bn_rxcuis;
drop if exists materialized view "Formularies".rxcui_to_brand_through_uspdc cascade;
create materialized view "Formularies".rxcui_to_brand_through_uspdc AS create materialized view "Formularies".rxcui_to_brand_through_uspdc AS
select select distinct
rccl.source_rxcui rccl.source_rxcui
,rccl.linked_rxcui ,rccl.linked_rxcui
,rccl.category ,rccl.category
,rccl."class" ,rccl."class"
,rr.tty1 ,rr.tty1
,rr.tty2 --,rr.tty2
,rr.rxcui2 ,rr.rxcui2
from "Formularies".rxcui_category_class_links rccl from "Formularies".rxcui_category_class_links rccl
join rxnorm_migrated.rxnorm_relations rr on rr.rxcui1 = rccl.linked_rxcui join rxnorm_migrated.rxnorm_relations rr on rr.rxcui1 = rccl.linked_rxcui
where rr.tty2 = 'BN' where rr.tty2 = 'BN'
; ;
/* So this one takes each RXCUI and it's associated RXCUIs from the same
category and class, and filters it down to associated RXCUI's that
represent brand names.
*/
create materialized view match_trial_compound_to_alternate_bn_rxcuis as create materialized view "Formularies".match_trial_compound_to_alternate_bn_rxcuis as
select distinct mttbi.nct_id, rtbtu.rxcui2 select distinct mttbi.nct_id, rtbtu.rxcui2 as brand_rxcuis
from match_trials_to_bn_in mttbi from match_trials_to_bn_in mttbi
join "Formularies".rxcui_to_brand_through_uspdc rtbtu join "Formularies".rxcui_to_brand_through_uspdc rtbtu
on mttbi.bn_or_in_cui = rtbtu.rxcui2 on mttbi.bn_or_in_cui = rtbtu.rxcui2
;
/* /*
Now I need to create a way to link This takes the list of ingredients and brands associated with a trial, and
links it to the list of alternative brand names.
*/ */
--renamed the view
CREATE OR REPLACE VIEW "Formularies".nct_to_brand_counts_through_uspdc
AS SELECT mtctabr.nct_id,
count(*) AS brand_name_counts
FROM "Formularies".match_trial_compound_to_alternate_bn_rxcuis mtctabr
GROUP BY mtctabr.nct_id;

128
logs.org
Unescape Escape View File

@ -0,0 +1,128 @@
* Plan/Todo [2025-01-06]
Goal is to update the main images with more details, i.e. adding means
etc.
- get aact_db back up
- attach it to a "research" network
- restart rocker, attaching it to the same research network.
** NOTES
aact_db-restored-2024-11-27 didn't successfully restore. It is missing
all the important stuff.
Figured out why the restore was failing. My code to restore had a faulty
check to see if the DB was up and ready. Fixed that now.
Waiting for restore (manually triggered) to start. Then I should have
access to the table as needed.
It seems like I'm missing some data within a schema, specifically the
Formularies and their associated views.
My options are:
- search around for documentation or other stuff
- try to rebuild
my suspision is that I forgot to back it up. I think it is probably
worth looking for. - So I've been looking through my copy of
ClinicalTrialsDataProcessing, and have not found anything referencing
it. The formularies data is required for my analysis though. If I
remember correctly, I manually uploaded the USP datasets in DBeaver,
then created any views etc.
I think that I'll have to recreate it. This is going to be hard because
I'm not sure what it did. At least I created mildly informative table
names.
The tables/views I've identified are: -
=Formularies.nct_to_brands_through_uspdc=
It looks like I need to - import usp-dc dataset - link those drugs to
usp data - create a view that links those automatically - back it up. -
double check the data I get from the request.
The links will be through RXCUIs, and grouped on =USP Class= In effect,
for a given RXCUI, I want to get the list of RXCUI's which have the same
USP-DC class, and then be able to link back to brands.
This should have the following links: - RXCUI -> USP-DC category/class
pair - USP-DC category/class pair -> RXCUIs - RXUCIs -> competitors
Do I want to combine the USP-DC and UPS-MMG datasets? No, there is
enough difference in them that I don't want to have to handle it that
way.
I've been working on this in scripts/ConfiguringFormularies.sql
So what I've managed to do so far is export tables, backup the data.
I've got a version that connects trials to brand names, but there may be
more details to the connection than I thought. I'd like to check if I
need to filter anything or check if there are other ingredients etc that
I need to include. */I probably need to write some descriptions of all
the tables and views to put everything together. An ai would probably be
helpful in doing this./*
** Code snippets
#+begin_example
podman run \
-e POSTGRES_PASSWORD="${POSTGRES_PASSWORD}" \
-e POSTGRES_USER="${POSTGRES_USER}" \
-e POSTGRES_DB="${POSTGRES_DB}" \
--name "${CONTAINER_NAME}" \
--detach \
--network research-network \
--shm-size=512mb \
--volume ./backup/:/backup/ \
-p 5432:5432\
postgres:14-alpine
#+end_example
#+begin_example
function check_postgres {
podman exec -i "${POSTGRES_DB}" psql -h localhost -U "${POSTGRES_USER}" -d "${POSTGRES_DB}" -c '\q' > /dev/null 2>&1
}
#+end_example
** Notes at end of day
- was reasonably productive in getting stuff unblocked for finishing
JMP, which i'll need to do before I leave town next week.
** What I've got to do tomorrow
I've got a version that connects trials to brand names, but there may be
more details to the connection than I thought. I'd like to check if I
need to filter anything or check if there are other ingredients etc that
I need to include. */I probably need to write some descriptions of all
the tables and views to put everything together. An ai would probably be
helpful in doing this./* At the end of it all, I should be able to get a
count of competing drugs per trial.
Once that is done, I can relink aact_db and rocker, then rerun my
analysis. Then I can adjust the images that I need for my JMP.
* [2025-01-07 Tue 12:01] notes
So what I've got to do is
** DONE Investigate what compounds are showing up in my current list
if that is what I want, then I'll be able proceed with redoing my images
if not, then I'll have to work on adjusting the views etc that I have.
I've looked through it and it seems to correct.
*** [[/mnt/will/large_data/Research_large_data/ClinicalTrialsDataProcessing/Scripts/ConfiguringFormularies.sql][ConfiguringFormularies.sql:81]] [2025-01-07 Tue 13:24]
I've tweaked these three views to make them clearer.
I also renamed the view of interest to ="Formularies".nct_to_brand_counts_through_uspdc= to better represent what it does.
** TODO Rerun the analysis
So it looks like I'll need to
1. take a backup of aact_db
2. restore from backup, putting the container in the research network
3. then rerun the analysis.
** TODO Add more details to images
This should include means and some quantiles
Write Preview
Loading…
Cancel
Save