More OBCI examples desired

Mar 20, 2023

Darren:

The latest OBCI has a lot of terms added, but I'd very much like to have examples for various classes that have nothing, so that I can make sure the modeling works for everything we might encounter. I'm interested in examples of the following (2-3 examples each should do it). Note that some of these can be combined (for example, a monitoring biomarker that results from a glycan measurement).

1) physiological characteristic biomarker - things like blood pressure perhaps? 2) glycan biomarker 3) transcript biomarker - something based on measuring transcripts. Is this done? 4) activity biomarker - I saw in LOINC that there are some tests that measure activities. These obviously measure things like outputs, but there can be cases where the activity is what is talked about. 5) monitoring biomarker for disease 6) monitoring biomarker for exposure 7) response biomarker 8) predictive biomarker - actual BEST predictive, not what many papers call predictive (a mistake often made in the OncoMX set) 9) safety biomarker 10) biomarker of exposure 11) any biomarker of exposure to "immaterial" things, like radiation 12) biomarker of phenotypic abnormality 13) more genetic variation biomarkers 14) radiographic biomarkers (aka image) 15) histological biomarkers - I can't get a handle on what these are. Some describe them as things like slides or tissue slices, others say it's cells (but what about flow cytometry?)

Some of these can come from the OncoMX set, but they need to be vetted. Almost every one I checked had some issue, to the point that I don't feel comfortable doing anything automated with it. Actually, that's good advice for any source, as I've seen errors in MarkerDB as well. I'd also like pointers to biomarker resources that can be processed automatically, particularly if they have the following information:

1) the change (increased, etc) 2) the entity 3) the biospecimen - especially in cases of genetic variation (what precisely gets tested?) 4) the disease/phenotype 5) the application - that is, used for diagnosis, monitoring, etc (the BEST category). I've seen statements, for example, to the effect that all genetic variation biomarkers are diagnostic, prognostic, and risk. Can that be verified?

I'm afraid I don't have the bandwidth to address all of these myself, so any help will be appreciated. I think once we have a decent representative sample, we can start moving toward the next steps.

Raja:

Yes, we can do this

Update 6/16/2023:

Darren received examples which Dan provided. It was an extensive list but not detailed in the OBCI format. I am working on providing examples that will be in that format so Darren can go through them.

Update 7/27/2023

Daniall: I included some notes in the notes section with some of my thoughts and I would like to know what you think as well. I tried to get an enzymatic biomarker, predictive biomarkers, physiological biomarker, and a biomarker of exposure as well. I reasoned through why they could be those but would want to see what you think. (file attached)

biomarker examples.xlsx

Darren: Thank you for the examples. I'm in the middle of something right now so I'll be able to give more thoughts tomorrow. For now, I'll try to give a distinction between an enzyme as assessed entity vs enzyme activity as assessed entity, and where to find definitions for the BEST biomarker types.

The definitions are easy - they are in the ontology file. As the definitions given there are the same as the BEST resource, you can also look at https://www.ncbi.nlm.nih.gov/books/NBK338448/def-item/glossary.biomarker/ and (for details) https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5813875/. Note especially how 'predictive' biomarkers are defined; these are ONLY for exposures to chemicals or drugs.

The enzyme vs activity is a nuanced difference. Measure of an enzyme can be done by either measuring the entity itself (protein or complex) or by measuring its activity. The latter aspect itself has two possibilities: (1) measuring the level of reaction product(s) or input(s)s; or (2) measuring the activity using some "black box" method. I'm interested in that last one, as these are the most difficult to find. These will be cases where--for example--some kit does the work and simply reports in terms of activity (perhaps as a percentage against normal?). In these cases we don't actually know what the material entity is.

Daniall: Will continue to work on these examples using this new information. Set up meeting with Darren soon to talk this out and hear more of the explanation

Update July 31, 2023:

Darren: aspartate transaminase - I looked into what gets measured, and it's always stated that it's 'level of AST'. To me the wording implies a measurement of the actual entity, like what might be determined using quantitative immunoblotting or some method that produces quantity of material. However, when I looked at how the measure is reported, it's given as international units, which are measures of activity. But these units are also defined with respect to quantity, which is to say that activity and quantity can be interconverted. Thinking further on it, and with awareness that measurement techniques can change, I think we need to treat material quantity and activity as somewhat equivalent. That is, in contrast to what I originally thought might be a new assessed entity type, 'activity', I think we could roll that into the relation indicated_by_difference_in_level_of (OBCI:1000014) and its children. Right now that relation explains that 'level' can refer to raw counts, percentages, ratios, and concentrations. I can add 'activities' to this.

That being said, in this case we know the nature of the entity that is being measured (a protein) even if we are measuring it using activity as a proxy. We should be on the lookout for cases where we don't know this. I guess this is the sort of example I'm looking for for 'activities'; those where the assessed entity isn't something necessarily material. Right now something like "brain wave activity" came to mind. For example, it could be 'decreased brain wave activity' indicated_by_below_normal_level_of 'brain wave activity', but I'm not sure what the assessed entity type would be. More examples like this would be useful.

HER2/neu protein - The biomarker_status (which I presume is simply what we call 'biomarker') should be 'increased expression of HER2/neu protein'. It is indeed a predictive biomarker, though not for breast cancer. Rather, elevated expression in breast cancer patients will help predict the effectiveness of therapies for those patients.

atrial fibrillation - This would be a risk biomarker based on the note you gave, not predictive (which assesses likelihood of a response to therapy). That is, presence of afib means there is an increased risk of having a stroke (actually having the stroke is not part of the definition of risk biomarkers). I note that you left the specimen type blank. This is good, because that means it MIGHT be an example of what I'm thinking would be a physiological biomarker: biomarkers for which there's no real specimen, or that arise from system-level considerations, or that are themselves medical conditions of interest. I often consider 'increased blood pressure' to be a good candidate, but since it has all three characteristics given above it doesn't help narrow things down. More examples like this would be useful.

3-phenoxybenzioc acid - from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5129381/, "The urinary metabolite, 3-phenoxybenzoic acid (3-PBA), is commonly used to assess human exposure to a number of pyrethroids." This tells you that it is a pharmacological/response biomarker, and that the exposure agent of interest is one of the pyrethroids (ingredient in some insecticides). That's the least we can say with confidence. I'm less certain about the Parkinson's connection, but if there was one that would make 3-PBA a risk biomarker, as you stated. However, the assessed biomarker entity type would be molecular--specifically, a small molecule biomarker.

clinical-biomarkers / OBCI

More OBCI examples desired #8

Darren:

Raja: