OHDSI Home | Forums | Wiki | Github

OMOP and biobanks

Hi,

I work within the world of biobanks/cohorts and clearly clinical annotation is a big area and therefore we have been looking at OMOP and the benefits it brings in this area.

We have noticed though that the representation of samples in OMOP is based heavily on their use to generate data.

Do you think there would be any interest in exploring how the European standard for biobank samples (MIABIS) could be integrated within OMOP?

We have several people from this community interested to engage and wondered the best way to do this?

Hi Philip,
Adding MIABIS to the OMOP vocabularies strikes me something that might have great benefit to the community. Posting this question to the OHDSI forums was the right way to start a conversation about that. Once there is clarity about the value of adding MIABIS, the process of getting it added to the OMOP vocabularies is straightforward.

Here is a link to the Wiki page for OHDSI work group that oversees the OMOP CDM and its vocabularies. This page describes the process for proposing modifications. Basically it involves creating an issue on this github site. Proposals with clearly defined set of use cases have a much stronger chance of acceptance. @Christian_Reich and @clairblacketer are people with senior responsibilities in CDM and vocab work group. Communicating directly with them about how to make the case effectively and any deadlines that are affecting your work in this area is probably a good idea. I hope you move this forward. Good luck.
Andrew

.

Apart from the vocabulary standards, there also needs some work on the implementation of samples in the CDM.

Currently samples are stored in the specimen table. However, this might not be ideal for biobanks. For instance, a direct relation between measurements and specimens is not present. In addition, the OHDSI tooling (Atlas) is not geared for searching through samples directly. Cohort selection always goes via persons. e.g. a question like ‘How many samples are there with characteristic x and y?’ cannot be done through the current tooling. (please correct me if I am wrong!)

The good news is that innovations are, in general, always well received by the OHDSI community. So my suggestion would be to set up a working group around the topic of samples/biobanks.

Thank you @MaximMoinat and @Andrew for your replies. This is encouraging.

What we are witnessing in the biobank arena is a ever increasing reliance on clinical attributes, i.e. people are searching for the cohort of people first and then samples are just another factor in that decision. We also have witnessed via our community work that when people select a biobank, one of the biggest influences in that decision is the data available.

So a query would be very much along the lines of, we want a cohort of people who have X,Y,Z and also a blood sample available.

There are a group of people from the biobank community willing to put in an effort and for info, our biobank search tool we have utilised OMOP for the clinical characteristics and then MIABIS alongside it. But it would be great if these were together.

Interesting. Could you please post few links that describe what is MIABIS.
I was only able to find some info in this github repo here
https://github.com/MIABIS/miabis/blob/master/machinereadableimplementation/databaseimplementation/ERD-MIABIS2.png

(not sure it is is current). Also, when v2 superceeded v1? Who is using MIAIS at the moment.

The most detail would be here:

It is the data standard that underpins the work in Europe for biobanks, via the EU wide infrastructure BBMRI-ERIC. The BBRMI Directory has about 100 million samples in it (although in aggregated form). The use of MIABIS does go beyond that but it is more a standard for the exchange and discovery of data about biobanks rather than a standard that people would use in their day to day systems.

And when I talk about this, I am not really thinking that MIABIS needs to be incorporated into OMOP as in the whole standard. But if you look at:

vs

they are pretty close.

My ideal therefore would be to have some minor mods and terminology inclusion to OMOP which could then mean you were able to generate a MIABIS profile from OMOP CDM.

Probably, the only pieces that are a fit are the ones that are patient-centric: “sample” and “samplecollectionattribute”. Something like that. The documentation is a little sloppy, but hey, who is complaining!

All the rest is really metadata. Not sure we need that in the OMOP.

Do you have an implementation, so one could take a look?

Hi @Christian_Reich, completely.

I think the power comes from adding some elements of the sample to OMOP and from that a MIABIS profile can be created. Rather than importing all of MIABIS into OMOP. The two that you picked out would be the starting point.

What is the process to proceed? Would a quick TC/VC be the best approach so that it can be understood what the proposal would be? Or is it easier if I go straight to the templates in git?

t