Review human annotations for ALB UniProtKB:P02768

[cmungall]

http://amigo.geneontology.org/amigo/gene_product/UniProtKB:P02768

Albumin has many annotations, but do these truly reflect the actual function? And it seems that we have no annotations to its actual function which is regulation of oncotic pressure (GO:0030104), and as a carrier protein for specific substances?

I'll review some annotations on a paper by paper basis. This is not comprehensive, and my reading of some papers may be too cursory.

• [x] ## Shotgun proteomics reveals specific modulated protein patterns in tears of patients with primary open angle glaucoma naïve to therapy.

http://amigo.geneontology.org/amigo/reference/PMID:23580065

All annotations are high-throughput. Either to 'retina homeostasis' or 'extracellular space of lacrimal gland'. I think the first is dubious and the second is not meaningful.

There are other high-throughput papers annotated, and these also produce not very meaningful annotations

• [x] ## Albumin prevents mitochondrial depolarization and apoptosis elicited by endoplasmic reticulum calcium depletion of neuroblastoma cells

http://amigo.geneontology.org/amigo/reference/PMID:16153637

yields annotations to negative regulation of apoptotic process + maintenance of mitochondrion location

From the abstract These results suggest that albumin protects cells from entering into apoptosis by preventing mitochondrial depolarization

Why not annotate to the more proximal and direct function (preventing mt depolarization, i.e. GO:0051902)?

(arguably the ontology could have the causal relation between GO:0051902 and apoptosis)

• [x] ## Human serum albumin and its N-terminal tetrapeptide (DAHK) block oxidant-induced neuronal death.

http://amigo.geneontology.org/amigo/reference/PMID:14726550

yields annotations to copper ion binding, antioxidant activity, and negative regulation of programmed cell death

The antioxidant activity seems justified, and copper ion binding is not wrong, but IMO unsatisfying as these two annotations should be combined together, is AFAIK the N-terminus of ALB has a function binding of free transition metals controlling their reactivity and limit their availability for the Fenton reaction.

see e.g. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3577569/

But I question the neg reg of PCD. It's certainly causally upstream of neuronal death, in the experimental system chosen. But regulation of a program is too strong.

If we really believe this annotation then why don't we just add a negatively regulates link in the ontology between antioxidant activity and apoptosis?

• [ ] ## all those binding annotations

We have lots of binding annotations. If the binding reflects a meaningful function, what is it? GO doesn't say. This is a well studied protein so this doesn't reflect a knowledge gap, just excessive annotation to the experiment.

• [ ] ## From reactome

Lots of annotations to 'extracellular' from Reactome, but not to the actual function which is transport, e.g. http://www.reactome.org/content/detail/R-HSA-264679 -- maybe we can look at next week @ukemi

[ValWood]

Hi @sylvainpoux could you assign to somebody at UniProt to reivew?

[deustp01]

Lots of annotations to 'extracellular' from Reactome, but not to the actual function which is transport, e.g. http://www.reactome.org/content/detail/R-HSA-264679

Good point. Albumin is a ubiquitous low specificity binder of all sorts of hydrophobic stuff under physiological conditions, but it would be appropriate, in the spirit of the "all those binding annotations" comment, to restrain ourselves and annotate only those cases i) where the binding has the effect of sequestering something that could do damage or be damaged or lost if it were left to circulate on its own; and / or ii) where the binding enables the transport of a cargo too water-insoluble to get to where it needs to be on its own. And in those cases, we could use more specific MF terms rather that generic "binding". We are starting to look for and fix cases like this as part of the project to get more complete GO-CAM models out of Reactome pathways (plain binding reactions yield incomplete GO-CAMs by definition), so if someone sees a way to detect these improve-able binding annotations systematically, that would be great. Otherwise, we will stumble on them as we traverse our incomplete GO-CAMs and fix them as we find them. @ukemi anything to add?

[pgaudet]

Shotgun proteomics reveals specific modulated protein patterns in tears of patients with primary open angle glaucoma naïve to therapy.

http://amigo.geneontology.org/amigo/reference/PMID:23580065

Delete BP annotations to GO:0001895 | retina homeostasis

[pgaudet]

Albumin prevents mitochondrial depolarization and apoptosis elicited by endoplasmic reticulum calcium depletion of neuroblastoma cells

http://amigo.geneontology.org/amigo/reference/PMID:16153637

Changed annotation to GO:0051902 as suggested

[pgaudet]

Albumin prevents mitochondrial depolarization and apoptosis elicited by endoplasmic reticulum calcium depletion of neuroblastoma cells

http://amigo.geneontology.org/amigo/reference/PMID:16153637

Removed annotation to neg reg of PCD

I dont think we can add a negatively regulates link in the ontology between antioxidant activity and apoptosis; I think there are responses to oxidative stress that are not apoptosis

[pgaudet]

I fixed the UniProt annotations, except for binding - this protein is not any worse than others in that respect. We need a more global discussion for binding.

I leave this open in case @deustp01 wants to do more reviews for Reactome (otherwise, @deustp01 , please close).

[deustp01]

To try to review all of the Reactome annotations to P02768 ALB, I went to the UniProt page for the protein, followed the link there to the complete list of GO annotations for ALB and found 19 from Reactome. All of these are assignments to CC terms - no MF or BP annotations at all. Also, except for one pair where we annotated a vesicle degranulation event that moved ALB from an intracellular location to an extracellular one generating two annotations, every annotation represents the results of distinct experiments from different publications. Indeed, as suggested by the last column that I added to the table of annotations, we are mostly annotating the ability of extracellular albumin to reversibly bind a variety of small molecules to facilitate their transport in the body, so the multiple similar annotations are not duplicates. (Which doesn't keep them from distorting overrepresentation analysis results, of course.) (But also note that the Reactome annotations of ALB with "binding" MF terms are not being picked up, at least by UniProt, so that issue raised by @cmungall above appears to have been resolved.) So, indeed, time to close the ticket. but @pgaudet I appear not to have power to close tickets. QuickGO-annotations-1693406187189-20230830.xlsx

[ValWood]

Thanks @deustp01 I hope that didn't take too long. I'm trying to get on top of the old annotation tracker tickets. Some are over a decade old!