Biosample geolocation data capture proposal

[v-rocheleau]

This is a description of an extension we implemented in order to capture geolocation information in Phenopacket Biosamples.
Our use case is for non-human datasets, but could just as well be used on human studies where this information is relevant.

Geolocation use case slides to be presented at the next GA4GH Phenopackets WG meeting

Problem statement

With human data, biosamples are taken in a clinical context in an hospital or clinic.
On the other hand, with non-human datasets the sample collection process involves a lot of leg work.

As an example, take a study where biologists are collecting data on a population of wild animals.
Researchers may collect samples directly from animals, or from what they leave behind (droppings, fur, etc).

Samples may be collected over very large areas, and over long periods of time.
The location of collection for the samples can be very important to capture, as well as the time of collection.
Also, being able to capture geolocation would be of great interest for environmental samples like eDNA.

Currently, Phenopacket does not support the capture of geolocation data in any way.

Proposal

Add a location_collected field to Biosample, analogous to time_of_collection but for the spacial dimension.

• Type GeoLocation (SchemaBlocks)
  • Allows the capture of a GeoJSON "Point" object for precise lat/long coordinates
  • Allows the capture of human-readable location descriptions
• Multiplicity: 0..1
  • This field should only be used when needed by the study
  • A sample is only collected at one location

We have already implemented this proposed schema extension in our platform, since this is a requirement for environmental and agricultural studies we need to support in the future.

We believe the proposed change is non-breaking and generic enough to benefit others in the Phenopackets community.

The GeoLocation schema block uses the widely adopted GeoJSON standard to represent coordinates, making it interoperable with most geo-information-system (GIS) tools.

Although our use case is non-human, GeoLocation (SchemaBlocks) seems to have been made with human data in mind as well.

[pnrobinson]

Thanks!
One thing to think about is that in medicine we are often more interested in the environmental exposures over the course of somebody's life. Consider the case of a farm work exposed to some pesticide for 20 years who travels 300 miles to the Mayo clinic where a biopsy is performed. We should discuss whether this element should be located in Biosample?

[v-rocheleau]

Hi Peter!

I agree with you, although I want to point out that this proposal is not intended to capture exposure data but to simply give a means to capture where a Biosample was collected. So the farm worker example doesn't apply to our use case right now, but I can see it come up eventually.

Exposure

Disclaimer for what follows: I have no exposure studies background, and have not been involved with the new exposure group yet, so please correct me if I'm wrong.

I don't think that the Biosample element is the right place to capture exposure data, I believe that the two are separate concerns.

Here are some ideas on how to capture exposure data, assuming the farm worker example.

Using measurements

Using the current Phenopacket schema, exposure data could be stored as Measurement elements.
Assuming that an individual's exposure is something that can be captured as a quantitative, ordinal or categorical measurement.

But I can see this being a bit too convoluted, and it would still require a way to add geolocation information.

Using dedicated exposure element

A better way to organize exposure data may require a dedicated element. Here is my humble proposal, making assumptions.

Say we use phenopacket.exposure as the top-level field for this:

• Multiplicity: 0..*
  • Exposure data is optional for phenopackets
  • Multiple exposure values can be given to a Phenopacket
• Type: Exposure
  • New data type for dedicated exposure data capture
  • Fields:
    • type: OntologyClass the type of exposure (environment, lifestyle, etc)
    • exposition: OntologyClass what the patient was exposed to (pesticide, radiation, tobacco, etc)
    • time_of_exposure: TimeElement
      • Timestamp for spontaneous exposure
      • Timerange for prolongued exposure (20 years for our farm worker)
    • exposition_level: Value | ComplexValue
      • Provides a way to measure the exposition level if needed
      • e.g. PPM, decibels, becquerel
    • location_of_exposure: GeoLocation
      • Where the exposure occured
    • response: ExposureResponse
      • New element to describe patient's biological response to the exposure
      • I'm assuming this should have a TimeElement to capture a delayed response
      • (going on a limb) Maybe even a GeoLocation if the response comes after a change of environment?
    • [Other fields deemed relevant by the exposomics community]