Two discrepancies in MOA `pert_iname` between samples and drugs files about lincs-cell-painting

[niranjchandrasekaran]

Ah you are right; I take that back. There I no InChIKey in the drug file and I have been using pert_iname to merge the two lists. This means, I have not been annotating these two compounds with their respective MOAs/targets. Thanks for bringing this to my attention.

[gwaybio]

cc @niranjchandrasekaran @shntnu - in case either of you have run into this issue and solved it before

[niranjchandrasekaran]

Thanks @gwaygenomics for figuring this out! It will be helpful when I am losing my mind over why the lists aren't matching.

On a side note: Usually, I merge the two lists (samples and drugs) using InChIKey. When Shantanu and I did that, we ran into another issue where there were multiple pert_iname mapping to the same InChIKey and InChIKey and smiles did not match 1:1. We have taken this approach to overcome the problems and the cleaned up samples file is here. This file also contains additional vendor availability information, which may not be useful to you.

Edit: I have been merging the two lists with pert_iname and not with InChIKey. But the rest of the paragraph is accurate for deduplicating the samples list.

[gwaybio]

oh interesting! Thanks for sharing @niranjchandrasekaran.

Where are you retrieving the drug file from? On the repurposing homepage in CLUE (https://clue.io/repurposing) the drug file does not have an InChIKey column.

This is the source of data in this repo (#5 (comment)) - maybe it is different in JUMP?

[gwaybio]

no longer an issue. Solved specifically in #8 (comment) and implemented in #7