Universally unique identifier (UUID) management

TheJacksonLaboratory / PDXintegrator

Code for mapping PDX Net data to a common data model

http://pdxintegrator.readthedocs.io/en/latest/

MIT License

1 stars 1 forks source link

Universally unique identifier (UUID) management #10

Open pnrobinson opened 6 years ago

pnrobinson commented 6 years ago

We need to decide upon a strategy for the various (patient, PDX model) identifiers to be used in PDX Net. Some of the issues include

What happens if there are multiple submissions from the same patient? Do we expect that the identifiers generated for PFX Net will remain identical for the same patient for a submission at two different time points? How does this get implemented?
Do the PDX centers also want to see the original IDs in searches?
How does the uuid related to uuids for the same patient in other databases if for example a sample from the same patient is processed for PDX Net and in another study?
Who is responsible for generating the uuids -- the centers themselves or PDXNet?
We should decide upon an SOP for this and seek consensus among the participating centers.

chuanglab commented 6 years ago

Carol and Steve have more direct experience with this, but here are my comments:

a tissue slide, tumor volume, sequencing data etc corresponds to a single sample, not a single patient. So there needs to be identifiers per sample, not just per patient.
there can be a lot of problems if the PDXnet uiud is based on the submitting group's id, e.g. if they give multiple samples from the same patient the same id. So PDXnet should generate the uiud.

sethi7ik commented 6 years ago

Agree with Jeff here (samples need UUIDs in addition to patients). The standard method for generating UUIDs are 32 character alphanumeric ids. An example of how to generate these UUIDs could be here: http://www.javapractices.com/topic/TopicAction.do?Id=56

It is trickier to figure out how to connect these IDs with IDs outside the PDXNet as the samples could be distributed from any center or from the PDMR. Also, differences in data usage consent on different studies can make it impractical for the same patient to have the same UUID on different studies.

However, I think that for PDXNet, we need to generate the UUID and give it to all the PDTCs.

pnrobinson commented 6 years ago

@chuanglab @sethi7ik who should be responsible for the topic of uuid's? I am not currently doing anything about this but I am concerned that nobody else is. Can we develop an SOP that we can present to the group at one of the next calls?

chuanglab commented 6 years ago

Carol developed a system such that the numbers are long enough to be unique but don't provide much biological info in themselves (e.g. a model is TM00099), and they had also been working on tracking info about related samples from the same patient. I believe Steve has worked on this with Carol.