bhklab / PharmacoDI_snakemake_pipeline

A Snakemake pipeline to automate the scripts for creating all PharmacoDB database tables and allow easy deployment on a range of platforms.
MIT License
1 stars 0 forks source link

UHNBreast Hill Slope values are NULL in the current PDB database #11

Open p-smirnov opened 2 years ago

p-smirnov commented 2 years ago

As an example, see this DDRC page: https://pharmacodb.ca/search?compound=Paclitaxel&cell_line=600MPE

If you inspect the request to PharmacoDBs graphql API, you find that for the UHNBreast experiment, the HS returned is Null. This explains why the curve is flat, and also explains why all UHNBreast curves are flat when you look here: https://pharmacodb.ca/search?compound=Paclitaxel&tissue=Breast

This could either be a PSet issue (misnamed column) or a PDI issue, but it should have been caught by PDI with an error message IMO.

ChristopherEeles commented 2 years ago

Issue exists in database but not in the UHNBreast PharmacoSet.

Likely due to the column name.

ChristopherEeles commented 2 years ago

Database query:

USE pharmacodb_test;

SELECT * FROM `profile` AS `p`
JOIN `experiment` AS `e`
ON `p`.`experiment_id` = `e`.`id`
JOIN `dataset` AS `d`
on `e`.`dataset_id` = `d`.`id`
WHERE `d`.`name` LIKE "UHNBreast";
ChristopherEeles commented 2 years ago

R code:

library(PharmacoGx)
library(data.table)

uhn <- downloadPSet("UHNBreast_2019")
sp <- sensitivityProfiles(uhn)
sp <- as.data.table(sp, keep.rownames=TRUE)
sp[is.na(slope_recomputed), ]
ChristopherEeles commented 2 years ago

R code has NA for only a subset of slope_recomputed but database query has NA for all values in the HS column.