Perturbation Metadata File - Perturbation ID and MOA

gwaybio commented 4 years ago

We have additional information for each compound assayed in the Drug Repurposing Hub Cell Painting Dataset.

There are at least four files on AWS, that could all work as a reference to describe compound metadata.

File Name	Columns
pert_info.txt	pert_id, pert_iname, pert_type, moa
pert_iname_moa.txt	pert_iname, moa, source, url, support, num_sources
pert_id_to_iname.txt	pert_id, pert_iname, pert_type
pert_iname_moa_aggregated.txt	pert_iname, moa, pert_iname_modified

Below I summarize each of the files

pert_info.txt

pert_iname_moa.txt

pert_id_to_iname.txt

pert_iname_moa_aggregated.txt

Confirmed pert_info.txt subset

gwaybio commented 4 years ago

We now need to decide how to report compound metadata information. In the past, i've used pert_info.txt (e.g. broadinstitute/cell-health#111). For completeness, my vote would be to combine pert_info.txt with information from the following columns: pert_iname_modified, source, url, support, and num_sources.

We can also add metadata for all compounds and add a boolean column (maybe cell_painted_A549) to denote which compounds we screened.

Also definitely open to suggestions and discussion!

gwaybio commented 4 years ago

cc @shntnu

shntnu commented 4 years ago

I'm looking up annotations on the Broad cluster

The annotations are going to be under the batch 2016_04_01_a549_48hr_batch1 but there are multiple folders corresponding to this batch because of the different ways we processed the data.

But there are only two folders 2016_04_01_a549_48hr_batch1 and 2016_04_01_a549_48hr_batch1_cmap_style that have annotation files. I compare the relevant files below.

metadata$ cd /cmap/imaging/2015_10_05_DrugRepurposing_AravindSubramanian_GolubLab_Broad/workspace/metadata
metadata$
metadata$ diff \
> ./2016_04_01_a549_48hr_batch1/barcode_platemap.csv \
> ./2016_04_01_a549_48hr_batch1_cmap_style/barcode_platemap.csv
metadata$
metadata$ diff \
> ./2016_04_01_a549_48hr_batch1/cell_painting_dataset_moa.csv \
> ./2016_04_01_a549_48hr_batch1_cmap_style/cell_painting_dataset_moa.csv
metadata$
metadata$ diff \
> ./2016_04_01_a549_48hr_batch1_cmap_style/cell_painting_dataset_cmap_annotations_moa.csv \
> ./2016_04_01_a549_48hr_batch1/cell_painting_dataset_cmap_annotations_moa.csv
metadata$

The README has an explanation of what's what

2016_04_01_a549_48hr_batch1$ cd /cmap/imaging/2015_10_05_DrugRepurposing_AravindSubramanian_GolubLab_Broad/workspace/metadata/2016_04_01_a549_48hr_batch1
2016_04_01_a549_48hr_batch1$ cat README
cell_painting_dataset_compound_list_anot.tsv    Annotation provided by Steven Corsello <corsello@broadinstitute.org>
cell_painting_dataset_moa.csv   Annotation obtained by querying Repurposing Hub (2016-12-08)
cpd_moa_and_gene_target_n90.txt Annotation provided by Lev Litichevskiy <lev@broadinstitute.org>
cell_painting_dataset_cmap_annotations_moa.csv Annotations copied from /cmap/projects/M1/annotation/pert_info_a2.txt

Note that these files were also shared with the PERISCOPE team here

Looking the run log for this dataset here, it looks like we annotated the profiles using cell_painting_dataset_moa.csv but then when running CMap's tools we decided to use the latter.

Side note: this is how cell_painting_dataset_cmap_annotations_moa.csv was created.

And this is how cell_painting_dataset_moa.csv was created i.e. by querying api.clue.io

So now the only thing to resolve is the difference between cell_painting_dataset_cmap_annotations_moa.csv and cell_painting_dataset_moa.csv. More on this next.

shntnu commented 4 years ago

Here is how they differ

First note that cell_painting_dataset_cmap_annotations_moa has only a subset of the annotations, so we can already conclude that between these two, cell_painting_dataset_moa is the more complete set.

na.omit(cell_painting_dataset_moa) %>% nrow()
#> [1] 1549
na.omit(cell_painting_dataset_cmap_annotations_moa) %>% nrow()
#> [1] 605
inner_join(na.omit(cell_painting_dataset_moa), na.omit(cell_painting_dataset_cmap_annotations_moa), by = c("pert_id", "pert_iname")) %>% nrow()
#> [1] 605
inner_join(na.omit(cell_painting_dataset_moa) %>% mutate(moa = tolower(moa)), na.omit(cell_painting_dataset_cmap_annotations_moa) %>% mutate(moa = tolower(moa)), by = c("pert_id", "pert_iname"))  %>% filter(moa.x != moa.y) %>% nrow()
#> [1] 146

^{Created on 2020-03-16 by the reprex package (v0.3.0)}

inner_join(na.omit(cell_painting_dataset_moa) %>% mutate(moa = tolower(moa)), na.omit(cell_painting_dataset_cmap_annotations_moa) %>% mutate(moa = tolower(moa)), by = c("pert_id", "pert_iname"))  %>% filter(moa.x != moa.y) %>% knitr::kable()

The two files disagree on 146/605 compounds. Some of these are differences are trivial (e.g. quinapril and ramipril. But most are of the category where only the first MOA listed was chosen.

I verified that for all compounds where the MOAs agree, there was only one MOA listed per compound

> inner_join(na.omit(cell_painting_dataset_moa) %>% mutate(moa = tolower(moa)), na.omit(cell_painting_dataset_cmap_annotations_moa) %>% mutate(moa = tolower(moa)), by = c("pert_id", "pert_iname"))  %>% filter(moa.x == moa.y) %>% pull(moa.x) %>% map_lgl(~str_detect(.x, "\\|")) %>% any()
[1] FALSE

For Broad's internal reference: we have a conversation about this in CMap's Slack channel here

pert_id	pert_iname	moa.x	moa.y
BRD-K25650355	physostigmine	cholinesterase inhibitor	acetylcholinesterase inhibitor
BRD-K26657438	imiquimod	interferon inducer\|tlr agonist	interferon inducer
BRD-K28542495	benzydamine	prostanoid receptor antagonist	membrane integrity inhibitor
BRD-K29359156	ebselen	cyclooxygenase inhibitor\|glutathione peroxidase agonist\|h+/k+-atpase inhibitor\|nitric oxide synthase inhibitor	cyclooxygenase inhibitor
BRD-K29582115	ziprasidone	dopamine receptor antagonist\|serotonin receptor antagonist	dopamine receptor antagonist
BRD-K29653726	topiramate	carbonic anhydrase inhibitor\|glutamate receptor antagonist\|kainate receptor antagonist	carbonic anhydrase inhibitor
BRD-K30480208	torasemide	electrolyte reabsorption inhibitor\|thromboxane receptor antagonist	electrolyte reabsorption inhibitor
BRD-K15916496	clotrimazole	cytochrome p450 inhibitor\|imidazoline ligand	cytochrome p450 inhibitor
BRD-K16444452	ibudilast	leukotriene receptor antagonist\|phosphodiesterase inhibitor	leukotriene receptor antagonist
BRD-A17883755	lenalidomide	anticancer agent	antineoplastic
BRD-K99964838	bosutinib	abl kinase inhibitor\|bcr-abl kinase inhibitor\|src inhibitor	abl inhibitor
BRD-K12184916	NVP-BEZ235	mtor inhibitor\|pi3k inhibitor	mtor inhibitor
BRD-K32744045	disulfiram	aldehyde dehydrogenase inhibitor\|dna methyltransferase inhibitor\|trpv agonist	aldehyde dehydrogenase inhibitor
BRD-K33211335	dextromethorphan	glutamate receptor antagonist\|sigma receptor agonist	glutamate receptor antagonist
BRD-A33447119	oxfendazole	anthelmintic agent	anthelmintic
BRD-A19195498	trimipramine	norepinephrine reputake inhibitor\|tricyclic antidepressant	norepinephrine reuptake inhibitor
BRD-K19687926	lapatinib	egfr inhibitor\|erbb2 inhibitor	egfr inhibitor
BRD-K25433859	maprotiline	norepinephrine reputake inhibitor\|tricyclic antidepressant	norepinephrine reuptake inhibitor
BRD-K07881437	danusertib	aurora kinase inhibitor\|growth factor receptor inhibitor	aurora kinase inhibitor
BRD-K08206212	entecavir	dna replication inhibitor\|reverse transcriptase inhibitor	dna replication inhibitor
BRD-K10843433	phenylbutazone	cyclooxygenase inhibitor\|prostanoid receptor antagonist	cyclooxygenase inhibitor
BRD-K10670311	sulfasalazine	antirheumatic drug\|nfkb pathway inhibitor	antirheumatic
BRD-K80738081	resveratrol	cytochrome p450 inhibitor\|sirt activator	cytochrome p450 inhibitor
BRD-K18895904	olanzapine	dopamine receptor antagonist\|serotonin receptor antagonist	dopamine receptor antagonist
BRD-A18917088	estradiol	estrogen receptor agonist	contraceptive agent
BRD-K19462402	buflomedil	adrenergic receptor antagonist\|calcium channel blocker	adrenergic receptor antagonist
BRD-A21858158	praziquantel	anthelmintic agent	anthelmintic
BRD-A24191444	ifenprodil	adrenergic receptor antagonist\|glutamate receptor antagonist	adrenergic receptor antagonist
BRD-K23984367	sorafenib	flt3 inhibitor\|kit inhibitor\|pdgfr tyrosine kinase receptor inhibitor\|raf inhibitor\|ret tyrosine kinase inhibitor\|vegfr inhibitor	flt3 inhibitor
BRD-K24576554	AT-9283	aurora kinase inhibitor\|jak inhibitor	abl inhibitor
BRD-A42759514	ornidazole	antiprotozoal agent	antiprotozoal
BRD-K44442813	pidotimod	interferon receptor agonist\|interleukin receptor agonist	interferon receptor agonist
BRD-A44448661	pentobarbital	barbiturate antiepileptic\|gaba receptor modulator	barbiturate antiepileptic
BRD-K51350053	toremifene	estrogen receptor antagonist\|selective estrogen receptor modulator (serm)	estrogen receptor antagonist
BRD-A51382177	fosinopril	angiotensin converting enzyme inhibitor	ace inhibitor
BRD-K52989797	clomipramine	serotonin transporter (sert) inhibitor	serotonin transporter inhibitor (sert)
BRD-K53737926	amitriptyline	norepinephrine inhibitor\|norepinephrine reuptake inhibitor\|serotonin receptor antagonist\|serotonin reuptake inhibitor	norepinephrine inhibitor
BRD-K54759182	dosulepin	norepinephrine reuptake inhibitor\|serotonin reuptake inhibitor\|tricyclic antidepressant	norepinephrine reuptake inhibitor
BRD-A56359832	zileuton	leukotriene inhibitor\|lipoxygenase inhibitor	leukotriene inhibitor
BRD-K67868012	PI-103	mtor inhibitor\|pi3k inhibitor	mtor inhibitor
BRD-K51033547	tramadol	norepinephrine reputake inhibitor\|opioid receptor agonist\|serotonin reuptake inhibitor	norepinephrine reuptake inhibitor
BRD-A68009927	daunorubicin	rna synthesis inhibitor\|topoisomerase inhibitor	rna synthesis inhibitor
BRD-K35960502	niclosamide	dna replication inhibitor\|stat inhibitor	dna replication inhibitor
BRD-K36927236	glyburide	atp channel blocker\|insulin secretagogue\|sulfonylurea	sulfonylurea
BRD-K92731339	perindopril	angiotensin converting enzyme inhibitor	ace inhibitor
BRD-A89585551	mefloquine	adenosine receptor antagonist\|hemoglobin antagonist	adenosine receptor antagonist
BRD-K90789829	nefazodone	adrenergic inhibitor\|norepinephrine reuptake inhibitor\|serotonin receptor antagonist\|serotonin reuptake inhibitor	adrenergic inhibitor
BRD-K31342827	GF109203X	pkc inhibitor	cdk inhibitor
BRD-K28428262	brivanib	fgfr inhibitor\|vegfr inhibitor	fgfr inhibitor
BRD-K32107296	temozolomide	dna alkylating drug	dna alkylating agent
BRD-A34255068	rolipram	phosphodiesterase inhibitor	interleukin receptor antagonist
BRD-A13084692	troglitazone	insulin sensitizer\|ppar receptor agonist	insulin sensitizer
BRD-K13646352	midostaurin	flt3 inhibitor\|kit inhibitor\|pkc inhibitor	flt3 inhibitor
BRD-K13819402	desoxypeganine	acetylcholinesterase inhibitor\|monoamine oxidase inhibitor	acetylcholinesterase inhibitor
BRD-K01638814	rilmenidine	adrenergic receptor agonist\|imidazoline receptor agonist	adrenergic receptor agonist
BRD-K06902185	minoxidil	katp activator\|kir6 channel (katp) activator\|vasodilator	katp activator
BRD-K01815685	indole-3-carbinol	aryl hydrocarbon receptor agonist\|indoleamine 2,3-dioxygenase inhibitor	aryl hydrocarbon receptor agonist
BRD-K02265150	amoxapine	norepinephrine reputake inhibitor	norepinephrine reuptake inhibitor
BRD-A02367930	ethinyl-estradiol	estrogen receptor agonist\|estrogenic component in oral contraceptives	dna directed dna polymerase stimulant
BRD-K02404261	caffeine	adenosine receptor antagonist\|phosphodiesterase inhibitor	adenosine receptor antagonist
BRD-A03216249	mepivacaine	potassium channel blocker\|sodium channel blocker	potassium channel blocker
BRD-K03384561	roquinimex	angiogenesis inhibitor\|tnf production inhibitor	angiogenesis inhibitor
BRD-K37289225	clozapine	dopamine receptor antagonist\|serotonin receptor antagonist	dopamine receptor antagonist
BRD-K37846922	3,3'-diindolylmethane	chk inhibitor\|cytochrome p450 activator\|indoleamine 2,3-dioxygenase inhibitor	chk inhibitor
BRD-K38436528	imipramine	norepinephrine reputake inhibitor\|serotonin reuptake inhibitor	norepinephrine reuptake inhibitor
BRD-K41260949	divalproex-sodium	gaba receptor agonist	hdac inhibitor
BRD-A43082555	loxoprofen	cyclooxygenase inhibitor\|prostanoid receptor antagonist	cyclooxygenase inhibitor
BRD-A44090213	indoprofen	cyclooxygenase inhibitor\|prostanoid receptor antagonist	cyclooxygenase inhibitor
BRD-A43882281	pinacidil	atp channel activator\|potassium channel activator	atp channel activator
BRD-K55966568	orantinib	fgfr inhibitor\|pdgfr tyrosine kinase receptor inhibitor\|vegfr inhibitor	fgfr inhibitor
BRD-A58955223	sulforaphane	anticancer agent\|aryl hydrocarbon receptor antagonist	antineoplastic
BRD-A39390670	rabeprazole	atpase inhibitor\|gastrin inhibitor	atpase inhibitor
BRD-K41170226	deoxycholic-acid	biliverdin reductase a activator\|g protein-coupled receptor agonist	biliverdin reductase a activator
BRD-A55913614	primaquine	antimalarial agent\|dna inhibitor	antimalarial
BRD-K57545991	enalapril	angiotensin converting enzyme inhibitor	ace inhibitor
BRD-K59273480	propentofylline	adenosine reuptake inhibitor\|phosphodiesterase inhibitor	adenosine reuptake inhibitor
BRD-K59369769	tozasertib	aurora kinase inhibitor\|bcr-abl kinase inhibitor\|flt3 inhibitor\|jak inhibitor	aurora kinase inhibitor
BRD-K60237333	niacin	nad precursor with lipid lowering effects\|vitamin b	nad precursor with lipid lowering effects
BRD-K51677086	erythromycin-ethylsuccinate	cytochrome p450 inhibitor\|protein synthesis inhibitor	cytochrome p450 inhibitor
BRD-A51714012	venlafaxine	adrenergic inhibitor\|norepinephrine reuptake inhibitor\|serotonin reuptake inhibitor	adrenergic inhibitor
BRD-K53318339	vinpocetine	phosphodiesterase inhibitor\|sodium channel blocker	phosphodiesterase inhibitor
BRD-K53857191	risperidone	dopamine receptor antagonist\|serotonin receptor antagonist	dopamine receptor antagonist
BRD-K54416256	methimazole	antithyroid agent	antithyroid
BRD-K07572174	curcumin	cyclooxygenase inhibitor\|histone acetyltransferase inhibitor\|lipoxygenase inhibitor\|nfkb pathway inhibitor	cyclooxygenase inhibitor
BRD-A48430263	pioglitazone	insulin sensitizer\|ppar receptor agonist	insulin sensitizer
BRD-A50311610	meclizine	constitutive androstane receptor (car) agonist	car agonist
BRD-A64977602	mirtazapine	adrenergic receptor antagonist\|serotonin receptor antagonist	adrenergic receptor antagonist
BRD-K63828191	raloxifene	estrogen receptor antagonist\|selective estrogen receptor modulator (serm)	estrogen receptor antagonist
BRD-K67277431	picotamide	thromboxane receptor antagonist\|thromboxane synthase inhibitor	thromboxane receptor antagonist
BRD-A48237631	mitomycin-c	dna alkylating agent\|dna synthesis inhibitor	dna alkylating agent
BRD-K48300629	zonisamide	sodium channel blocker\|t-type calcium channel blocker	sodium channel blocker
BRD-K49328571	dasatinib	bcr-abl kinase inhibitor\|ephrin inhibitor\|kit inhibitor\|pdgfr tyrosine kinase receptor inhibitor\|src inhibitor\|tyrosine kinase inhibitor	bcr-abl kinase inhibitor
BRD-K49865102	PD-0325901	mek inhibitor	map kinase inhibitor
BRD-K50422030	clomethiazole	gaba receptor antagonist\|gaba receptor modulator	gaba receptor antagonist
BRD-A50675702	fipronil	chloride channel blocker\|gaba gated chloride channel blocker	chloride channel blocker
BRD-K50398167	meclofenamic-acid	cyclooxygenase inhibitor\|prostanoid receptor antagonist	cyclooxygenase inhibitor
BRD-A83081521	finasteride	5 alpha reductase inhibitor	5-alpha reductase inhibitor
BRD-K86930074	cediranib	kit inhibitor\|vegfr inhibitor	kit inhibitor
BRD-A45889380	mepacrine	cytokine production inhibitor\|nfkb pathway inhibitor\|tp53 activator	cytokine production inhibitor
BRD-K47869605	podophyllotoxin	microtubule inhibitor\|tubulin inhibitor	microtubule inhibitor
BRD-K81528515	nilotinib	abl kinase inhibitor\|bcr-abl kinase inhibitor	abl inhibitor
BRD-K81473089	tacrine	acetylcholinesterase inhibitor	acetylcholine release stimulant
BRD-K76908866	CP-724714	egfr inhibitor\|receptor tyrosine protein kinase inhibitor	egfr inhibitor
BRD-K68103045	CGS-20625	benzodiazepine receptor agonist\|gaba benzodiazepine site receptor partial agonist	benzodiazepine receptor agonist
BRD-K70358946	aripiprazole	serotonin receptor agonist\|serotonin receptor antagonist	serotonin receptor agonist
BRD-K70778732	trazodone	adrenergic receptor antagonist\|serotonin receptor antagonist\|serotonin reuptake inhibitor	adrenergic receptor antagonist
BRD-K72222507	quinapril	angiotensin converting enzyme inhibitor	ace inhibitor
BRD-K68488863	ENMD-2076	aurora kinase inhibitor\|flt3 inhibitor\|vegfr inhibitor	aurora kinase inhibitor
BRD-K70557564	zosuquidar	p glycoprotein inhibitor	p-glycoprotein inhibitor
BRD-K71035033	masitinib	kit inhibitor\|pdgfr tyrosine kinase receptor inhibitor\|src inhibitor	kit inhibitor
BRD-A68281735	REV-5901	leukotriene receptor antagonist\|lipoxygenase inhibitor	leukotriene receptor antagonist
BRD-K68867920	quetiapine	dopamine receptor antagonist\|serotonin receptor antagonist	dopamine receptor antagonist
BRD-K35189033	levonorgestrel	estrogen receptor agonist\|glucocorticoid receptor antagonist\|progesterone receptor agonist\|progesterone receptor antagonist	estrogen receptor agonist
BRD-A70083328	secnidazole	acetylcholinesterase inhibitor\|microtubule inhibitor	acetylcholinesterase inhibitor
BRD-K71103788	duloxetine	norepinephrine reuptake inhibitor\|serotonin reuptake inhibitor	norepinephrine reuptake inhibitor
BRD-K74305673	IKK-2-inhibitor-V	ikk inhibitor\|nfkb pathway inhibitor	ikk inhibitor
BRD-A75172220	hydrocortisone	glucocorticoid receptor agonist	corticosteroid agonist
BRD-K78692225	leflunomide	dihydroorotate dehydrogenase inhibitor\|pdgfr tyrosine kinase receptor inhibitor	dihydroorotate dehydrogenase inhibitor
BRD-K78431006	crizotinib	alk tyrosine kinase receptor inhibitor	alk inhibitor
BRD-K91301684	noscapine	bradykinin receptor antagonist\|tubulin inhibitor	bradykinin receptor antagonist
BRD-K74514084	pazopanib	kit inhibitor\|pdgfr tyrosine kinase receptor inhibitor\|vegfr inhibitor	kit inhibitor
BRD-A75479906	rimantadine	antiviral\|rna synthesis inhibitor	antiviral
BRD-K75641298	metoclopramide	dopamine receptor antagonist\|serotonin receptor antagonist	dopamine receptor antagonist
BRD-K78126613	menadione	mitochondrial dna polymerase inhibitor\|phosphatase inhibitor	mitochondrial dna polymerase inhibitor
BRD-K79131256	albendazole	tubulin inhibitor	anthelmintic
BRD-K89348303	ramipril	angiotensin converting enzyme inhibitor	ace inhibitor
BRD-K89162000	tandutinib	flt3 inhibitor\|kit inhibitor\|pdgfr tyrosine kinase receptor inhibitor	flt3 inhibitor
BRD-K91315211	betahistine	histamine receptor agonist\|histamine receptor antagonist	histamine receptor agonist
BRD-K91601245	mercaptopurine	immunosuppressant\|protein synthesis inhibitor\|purine antagonist	immunosuppressant
BRD-A91699651	chloroquine	antimalarial agent	antimalarial
BRD-K99749624	linifanib	pdgfr tyrosine kinase receptor inhibitor\|vegfr inhibitor	pdgfr receptor inhibitor
BRD-K92723993	imatinib	bcr-abl kinase inhibitor\|kit inhibitor\|pdgfr tyrosine kinase receptor inhibitor	bcr-abl kinase inhibitor
BRD-K93880783	stavudine	dna directed dna polymerase inhibitor\|reverse transcriptase inhibitor	dna directed dna polymerase inhibitor
BRD-K95763993	trapidil	pdgfr tyrosine kinase receptor inhibitor	pdgfr receptor inhibitor
BRD-K96319534	phentermine	dopamine uptake inhibitor\|serotonin reuptake inhibitor	dopamine uptake inhibitor
BRD-K93034159	cladribine	adenosine deaminase inhibitor\|ribonucleoside reductase inhibitor	adenosine deaminase inhibitor
BRD-K92428153	mycophenolate-mofetil	dehydrogenase inhibitor\|inositol monophosphatase inhibitor	dehydrogenase inhibitor
BRD-A92537424	danazol	estrogen receptor antagonist\|progesterone receptor agonist	estrogen receptor antagonist
BRD-K92984783	melperone	dopamine receptor antagonist\|serotonin receptor antagonist	dopamine receptor antagonist
BRD-K93460210	lamotrigine	serotonin receptor antagonist\|sodium channel blocker	serotonin receptor antagonist
BRD-A93424738	dexamethasone-acetate	glucocorticoid receptor agonist	corticosteroid agonist
BRD-K93754473	tamoxifen	estrogen receptor antagonist\|selective estrogen receptor modulator (serm)	estrogen receptor antagonist
BRD-K94830329	ataluren	cftr channel agonist\|dystrophin stimulant	cftr channel agonist
BRD-K63750851	mycophenolic-acid	dehydrogenase inhibitor\|inositol monophosphatase inhibitor	dehydrogenase inhibitor
BRD-K97440753	dihydroergocristine	adrenergic receptor antagonist\|prolactin inhibitor	adrenergic receptor antagonist
BRD-A97437073	rosiglitazone	insulin sensitizer\|ppar receptor agonist	insulin sensitizer

gwaybio commented 4 years ago

interesting, it looks like moa.y (column moa in cell_painting_dataset_moa.csv) is truncated by | and also remapped based on some dictionary somewhere.

Recode	Example
by `pipe`	`dopamine receptor antagonist <pipe> serotonin receptor antagonist` becomes `dopamine receptor antagonist`
some internal dictionary	`glucocorticoid receptor agonist` becomes `corticosteroid agonist`

Before we decide on a master reference list, it would be good to understand some decisions that went into compiling this list. A couple of questions below:

Is this mapping provided anywhere? Perhaps at /cmap/projects/M1/annotation/pert_info_a2.txt?
Are these two recoding strategies equivalent? i.e. assuming that CMAP has provided us with absolute ground truth, do we expect that recoding "by |" loses equivalent information as reccoding by "some internal dictionary"?
Is the ordering of the moas between | meaningful? i.e. is it safe that the first one selected is the most representative?
Is presence/absence of a specific moa in a | list meaningful?
- Example: is the serotonin receptor antagonist entry in norepinephrine inhibitor|norepinephrine reuptake inhibitor|serotonin receptor antagonist|serotonin reuptake inhibitor different than the serotonin receptor antagonist entry in dopamine receptor antagonist|serotonin receptor antagonist?
- Does the length of the list mean anything? Like this compound is super non-specific, or something similar?

My vote would be to include all information, rather than recoding. Or, possibly, do both. We can include a column moa_recode_simple or something like that, provide instructions on how the column was formed, and include both columns. An additional thing to consider depends on the meaning of the pipes (|). We could split on the pipes and duplicate compound rows to make the table longer.

e.g.

Compound	moa	moa_recode_simple
Magic Compound A	`dopamine receptor antagonist <pipe> serotonin receptor antagonist`	`dopamine receptor antagonist`

becomes:

Compound	moa	moa_recode_simple	moa_expand
Magic Compound A	`dopamine receptor antagonist <pipe> serotonin receptor antagonist`	`dopamine receptor antagonist`	`dopamine receptor antagonist`
Magic Compound A	`dopamine receptor antagonist <pipe> serotonin receptor antagonist`	`dopamine receptor antagonist`	`serotonin receptor antagonist`

These things have probably already been thought deeply about. If a decision has already been made, then please disregard these musings! It will still be very good to zero in on a master list, so that many downstream projects can benefit

gwaybio commented 4 years ago

Noting important distinction here after zooming with @shntnu - the pipes (|) are just standard delimiters. They most likely just mean that the specific compound is all of the moas. It is also likely that the order between pipes is not relevant.

Although something else to possibly consider is strength of assignment - i.e. how confident are we that the specific compound is annotated with a specific moa? (maybe this is too complicated, and not really necessary though)

I am set to meet with the CLUE team over virtual office hours at 1pm tomorrow https://clue.io/office-hours

shntnu commented 4 years ago

All this comes down to resolving a question (Greg does not have access) I had back in Oct 2016: how do we explain the differences between the annotations that Steven Corsello gave us cell_painting_dataset_compound_list_anot.tsv (*) and those that were generated by querying api.clue.io cell_painting_dataset_moa.csv.

I think the best path forward to is to get the latest MOA annotations from CMap (and just double check that they have been "approved" by Steven Corsello) @gwaygenomics

(*) -- Steven's note when he sent me this file: Steven CorselloAug 31st, 2016 at 9:39 AM I joined in the compound names and MoAs. This required some manual curation as a few of the perts were not actually in the REP library (some were added to the L1000 plates for other reasons). Looks like we have a few MoAs with multiple compounds. See you in a bit!

shntnu commented 4 years ago

@gwaygenomics note that I have updated my comments above https://github.com/broadinstitute/lincs-cell-painting/issues/5#issuecomment-599520284 and https://github.com/broadinstitute/lincs-cell-painting/issues/5#issuecomment-599523129

gwaybio commented 4 years ago

Thanks @shntnu - this additional context is helpful.

I had a productive conversation with @tnat1031 at CLUE office hours this afternoon. I will summarize our conversation below (most of this is probably obvious 😄):

The CMAP team hosts drug repurposing hub data on CLUE. The CMAP team is not directly responsible for drug annotations - Steven Corsello's team provides the annotations.
- The CMAP team will update their resources, when Steven sends an update
Currently, the data are versioned on the CLUE resource (screenshot below) on their landing page.
- There are two main files repurposing_samples and repurposing_drugs. The samples file contains extended Broad IDs (containing batch and aliquot information) and the drugs file contains MOAs and targets (!)( 👈 that is awesome!)
- The files are versioned by date (currently 20180907 for both)
- (We should also probably version them by md5 hash in this repo, since date can get iffy)
The order of the | delimiter in the MOA file is important (it is not clear if this is also true for the target column)
- The first MOA listed has more literature support than the other moas. The other MOAs are "less supported"
- So, if two compounds have the same MOA, the order only matters if one of the compounds has the MOA listed first
In terms of resolving MOAs with different names, there is no one-stop solution, and this is an issue CMAP also deals with.
- The recommended approach would be to manually consolidate and retain a running list of compound synonyms in an external file

Next Steps

I think the next steps are as follows:

Download the two files from CLUE and add them to this repo somewhere like metadata/
Copy their access policy in the metadata/ folder README (pasted below)
Include version information (date and MD5 check) in that README
Add in a jupyter notebook file to merge the two files, stripping the headers, and making a single unified resource.
Add in a jupyter notebook to explore the data - ask questions like (how many compounds per MOA, how many targets, basically any interesting metadata question)
Settle on a processing pipeline to get the cell painting profiles online! (to be described in a separate issue)

DO I NEED TO REGISTER TO ACCESS INFORMATION FROM THE REPURPOSING HUB? No, the annotations provided in the hub are freely available for research use by any organization. The information in the Repurposing Hub may not be repackaged or redistributed for commercial purposes without permission.

shntnu commented 4 years ago

@gwaygenomics This notebook is not related to the MOA question but will be pertinent https://rpubs.com/shantanu/repurp-annotations which is the same as this notebook

shntnu commented 4 years ago

A relevant conversation thread:

Forwarded Conversation Subject: Repurposing annotations – duplicates?

From: Shantanu Singh

Hi Steven, I hope you are well. JT and I were using the repurposing collection
annotations for one of our projects, and I noticed that a small handful of
drugs have duplicate entries (same pert_ids, but different drug names, and
their MOA and target annotations don't always match) Here's a
summary:http://rpubs.com/shantanu/repurp-annotations

Do you have any advice on how to handle this? Is it reasonable to merge all
the annotations for a pert_id, as I have done in the notebook? ThanksShantanu

From: Steven Corsello

Hi Shantanu, Thanks for looking into this. Most of these issues are related to
incorrect structure/stereochemistry (or a mixture treated as a salt form) for
a registered compound. Some have been fixed (by Josh, cc'd) but we have not
generated a fresh export of the compounds IDs from CBIP for one year. The core
Broad ID changes when the structure is curated. We're planning a database
update for this fall that will also add ~800 new compounds to the library. For
now, I suggest trusting our assigned name and metadata over the structure/core
ID alone for these compounds. Josh, could you please check to see if all of
these structures are fixed? Best,Steven

From: Joshua Sacher 

Hi Shantanu, Thanks for checking through the data set and providing the
relevant code. This is actually a surprisingly small list compared to what
needs to be done, so I'm happy to go through it and update any data you need
by hand. The more interesting/difficult list is when 1 name has more than 1
Core ID (see Actinomycin-d, for instance). There will be ~400
curations/changes in the next update to begin to address this. What
information would be helpful for your experiments? The correct Broad ID for
each of these names? Josh

From: Shantanu Singh 

Hi Steven and Josh, Thanks for getting back to me so quickly.  The
inconsistencies that I found are those where the 1 core ID has multiple names
(as well as different MOAs).  I had initially assumed a different explanation
– that the multiple names are in fact synonyms of sorts, and that the
differences in the MOA annotations were just incomplete annotation. But it
looks like these are actually ambiguous mappings i.e. totally different drugs
that have the same core ID (a.k.a. pert_id), and that's what I'm hoping to
resolve. I have attached these cases (duplicate_annotations.csvsame as the
second table in the notebook). Within this set, I'm looking to resolve
annotations of 7 compounds that were tested in Cell Painting
(duplicate_pert_ids_and_broad_ids_cell_painting.csv; broad ids are included),
and the question here is: which compounds to these correspond to, given the
multiple names? In all, this a pretty tiny problem: just 7 out of 1500+
compounds we have tested in Cell Painting have this issue. So its really not a
big deal for us to drop those 7 in our analysis, but please keep me posted
when you fix it. Thanks again!

Shantanu

Notes for myself:cell_painting_compounds %>% inner_join(duplicate_pert_ids)
%>% select(-n) %>% inner_join(annotated_compound_list %>% select(pert_id,
broad_id) %>% distinct()) %>%
write_csv("duplicate_pert_ids_and_broad_ids_cell_painting.csv")

duplicate_annotations.csv.txt duplicate_pert_ids_and_broad_ids_cell_painting.csv.txt

shntnu commented 4 years ago

The first MOA listed has more literature support than the other moas. The other MOAs are "less supported"

Here's a list of compounds with multiple MOAs. It doesn't appear that this ^ is true because the MOAs are sorted lexicographically.

repurposing_annotations <- read_tsv("https://s3.amazonaws.com/data.clue.io/repurposing/downloads/repurposing_drugs_20180907.txt", comment = "!")

repurposing_annotations %>% 
  mutate(n_moas = 1 + str_count(moa, "\\|")) %>% 
  filter(str_detect(moa, "\\|")) %>% 
  arrange(desc(n_moas)) %>% 
  select(pert_iname, moa, n_moas) %>% 
  mutate(moa = str_replace_all(moa, "\\|", ";")) %>% 
  knitr::kable() %>% 
  write_lines("~/Desktop/moa.txt")

I get the same ordered via the API e.g. https://api.clue.io/api/rep_drug_moas/?filter={%22where%22:{%22pert_iname%22:%22BMS-777607%22},%22fields%22:{%22name%22:true}}&user_key=dd910094ca286fc4ae3b174ae0ca70b1

[{"name":"AXL kinase inhibitor"},{"name":"c-Met inhibitor"},{"name":"FLT3 inhibitor"},{"name":"hepatocyte growth factor receptor inhibitor"},{"name":"macrophage migration inhibiting factor inhibitor"},{"name":"tyrosine kinase inhibitor"}]

pert_iname	moa	n_moas
BMS-777607	AXL kinase inhibitor;c-Met inhibitor;FLT3 inhibitor;hepatocyte growth factor receptor inhibitor;macrophage migration inhibiting factor inhibitor;tyrosine kinase inhibitor	6
dasatinib	Bcr-Abl kinase inhibitor;ephrin inhibitor;KIT inhibitor;PDGFR tyrosine kinase receptor inhibitor;src inhibitor;tyrosine kinase inhibitor	6
regorafenib	FGFR inhibitor;KIT inhibitor;PDGFR tyrosine kinase receptor inhibitor;RAF inhibitor;RET tyrosine kinase inhibitor;VEGFR inhibitor	6
sorafenib	FLT3 inhibitor;KIT inhibitor;PDGFR tyrosine kinase receptor inhibitor;RAF inhibitor;RET tyrosine kinase inhibitor;VEGFR inhibitor	6
amuvatinib	FLT3 inhibitor;KIT inhibitor;PDGFR tyrosine kinase receptor inhibitor;RAD51 inhibitor;RET tyrosine kinase inhibitor	5
dovitinib	EGFR inhibitor;FGFR inhibitor;FLT3 inhibitor;PDGFR tyrosine kinase receptor inhibitor;VEGFR inhibitor	5
LY294002	mTOR inhibitor;PI3K inhibitor;DNA dependent protein kinase inhibitor;phosphodiesterase inhibitor;PLK inhibitor	5
sunitinib	FLT3 inhibitor;KIT inhibitor;PDGFR tyrosine kinase receptor inhibitor;RET tyrosine kinase inhibitor;VEGFR inhibitor	5
amitriptyline	norepinephrine inhibitor;norepinephrine reuptake inhibitor;serotonin receptor antagonist;serotonin�norepinephrine reuptake inhibitor (SNRI)	4
caffeic-acid	HIV integrase inhibitor;lipoxygenase inhibitor;nitric oxide production inhibitor;tumor necrosis factor production inhibitor	4

shntnu commented 4 years ago

Clarifying note from Ted:

Yes, I think in the REP hub the MoAs are lexicographically ordered. When I spoke to Greg on Tuesday, he was also using a file of MoA annotations that he (or maybe you) had gotten from CMap a while back, and in that file we had ordered them according to how much literature support each had. If it helps, I think if there's any discrepancy between the MoAs for a given compound, the REP hub data should be considered the authority.

gwaybio commented 4 years ago

@shntnu - in #7 I begin processing the CLUE repurposing data. One additional question: it looks like the pert_id column is not provided. Is this something I should add? I imagine it would simply be truncating the broad_id column (e.g. BRD-K89787693-001-01-1 becomes BRD-K89787693)

I can then also create a third file (something like repurposing_info_simple.tsv) that only has unique columns pert_id, pert_iname, moa, and target. This file would be useful, because, presumably pert_id is the primary mapping between the CLUE resources and the Cell Painting profiles. cc @niranjchandrasekaran

shntnu commented 4 years ago

@shntnu - in #7 I begin processing the CLUE repurposing data. One additional question: it looks like the pert_id column is not provided. Is this something I should add? I imagine it would simply be truncating the broad_id column (e.g. BRD-K89787693-001-01-1 becomes BRD-K89787693)

Yes, the first two components of the broad_id are the pert_id (defined here A unique identifier for a perturbagen that refers to the perturbagen in general, not to any particular batch or sample.)

I can then also create a third file (something like repurposing_info_simple.tsv) that only has unique columns pert_id, pert_iname, moa, and target. This file would be useful, because, presumably pert_id is the primary mapping between the CLUE resources and the Cell Painting profiles. cc @niranjchandrasekaran

This seems reasonable. Note that with JUMP CP, we ended up with a proliferation of such derived files (but that was because we were dealing with many sources), and it can get confusing after a while (ideally, we'd design it like a relational database). But this will likely be the only derived metadata file, so that's ok.

gwaybio commented 4 years ago

the perturbation metadata file is added in #7

Based on discussions with @shntnu , @niranjchandrasekaran , @jrsacher , and @tnat1031 we've settled on repurposing hub metadata resources and decided to add 3 perturbation metadata files to the lincs_cell_painting repo.

Thanks everyone for the speedy contributions and comprehensive documentation!

Below is a description of the three resulting output files after processing. I will close this issue now that it is addressed in #7

Metadata Resources Added

Filename	Description	Derivation	Dimensions
`repurposing_info.tsv`	The primary file storing every column in repurposing hub drugs and samples metadata file	Inner join on drugs and samples `pert_iname` column	13,553 x 17
`repurposing_info_long.tsv`	Stores every unique perturbation with singular values in `target` and `moa` columns	Split aforementioned columns by pipe (e.g. entry `ACT1<pipe>ACT3` becomes two independent rows)	39,471 x 20
`repurposing_simple.tsv`	Core identifiers and biological metadata	Splits off aliquot and batch information from Broad ID to create a `pert_id` and drops duplicate entries in `target`, `moa`, and `pert_iname` columns	6,806 x 4

TODO

Note, steps 5 and 6 outlined in https://github.com/broadinstitute/lincs-cell-painting/issues/5#issuecomment-600208689 still need to be addressed:

Add in a jupyter notebook to explore the data - ask questions like (how many compounds per MOA, how many targets, basically any interesting metadata question)

Settle on a processing pipeline to get the cell painting profiles online! (to be described in a separate issue)

I will document these steps in a different issue and PR

broadinstitute / lincs-cell-painting