The study deletion issue

[ElDeveloper]

Study deletion is tricky for many reasons, but we need to figure this out.

[ackermag]

I think we need to consider old studies where bugs still existed. I don't
believe this will be an ongoing issue

On Fri, Feb 20, 2015 at 10:02 PM, Yoshiki Vázquez Baeza <
notifications@github.com> wrote:

Study deletion is tricky for many reasons, but we need to figure this out.

—
Reply to this email directly or view it on GitHub
#879.

Gail Ackermann
Knight Lab
UCSD
glackermann@ucsd.edu ackermag@ucsd.edu

[antgonza]

This issue is related to: #596

I have started working on deleting studies and I do not see it as
difficult because I think to be able to erase a study, you will need
to have no sample templates (not yet implemented), no prep templates
(implemented), etc. I think the most difficult one is to delete
process data as then we will need to figure out if there are part of a
(meta)analysis, if not fine to erase but if they are we could: allow
if not public, send emails to all people that use that meta-analysis
and flag it (add a message). Other option is to do all those steps,
except that we do not flag and erase those too (I do not like this
idea).

BTW, in the last update to the database, @Josename and me, had to
clean all broke studies, sample and prep templates and now we have
queues to add these so we should not have broken studies anymore.

What do you think?

[josenavas]

As @antgonza states, the trickiest step is removing processed data. This is becoming important and I'm going to start working on deleting ProcessedData, as it is needed for some of the EMP studies that we currently have on the DB. Thus, it will be great if we can have all the devs (ping! @biocore/qiita-dev ) commenting in here so we reach consensus and I can start working on it.

This is how I think it should work. Allow removal if and only if:

• The processed data is not public (we already decided that what we are going to publish are processed data, rather than studies).
• The processed data has not been used in any meta-analysis (we can then think on implementing meta-analysis removal).
• An admin can remove a processed data in any case, and then send an email to the users that have meta-analysis with such study (only needed in case that there are IRB problems)

What others think?

[antgonza]

Sounds fine except that I'm not sure we should erase all meta-analysis (even by hand) that have used a certain about to be deleted processed data vs. flagging them and make it clear in the GUI that this is the case. What others think? Also, just to confirm: currently we couldn't remove a processed data that has been used in a meta-analysis as the DB will not allow it, right?

[josenavas]

Yes, you're right, the DB will not allow it. Since we have #855 which aims to add a status to the processed data, we can add a new status such like "pending for removal" and we then flag all the meta-analysis that need to be removed. Here we can be more flexible: if there are IRB problems, we need to remove those meta-analysis immediately, if not, we can put some type of countdown: i.e this meta-analysis will be removed in 7 days.

[squirrelo]

Because of the analysis_sample table I think you are correct. We can't remove the processed data if it is used in an analysis.

I agree with @josenavas implementation, although we do need a way for sandbox/private studies to be deleted even with analyses on them. I'm thinking we send an email to the users who used the study, and have a BIG RED BANNER that warns them on the analysis results page.

[adamrp]

Re: IRB concerns and removing things immediately, it might be a good idea to have yet another status, e.g., "Suppressed," where the data is no longer available/visible (except to admins), so that we can quickly suppress data without needing to go through the process of deleting it immediately. This is similar to what EBI and other archives do.

[josenavas]

Assigning this to Alpha 0.2 as what is discussed here is how to delete studies that are public or with IRB concerns.