Phenotype file is parsed wrongly. Update the code to be able to parse the new file:

http://compbio.charite.de/jenkins/job/hpo.annotations/lastStableBuild/artifact/misc/phenotype_annotation.tab

Send them an email.
But It would be best to add a parameter in the config file so the server admin can decide to notify contacts or no. Something like this:

MATCH_NOTIFICATIONS = True

It's OK to wait until something works!

Host for instance. And remove host as a param in external_matcher, because it's not used any more.
Totally NOT urgent, to do when there is time

Not really a bug here, but it would add more useful info to the match objects.

Match objects saved to database now look like this:

matches = [
        {    # External match where test_patient is the query and with results
            '_id' : 'match_1',
            'has_matches' : True,
            'data' : {
                'patient' : {
                    'id' : 'P0000079',
                    'contact' : {
                        'href' : 'mailto:[email protected]'
                    }
                }
            },
            **'results' : [
                {'patient' : { 'patient_data' : 'test_stuff'}},
                {'patient' : { 'patient_data2' : 'test_stuff2'}},
            ],**
            **'searched_nodes' : [{ 'id': 'node1_id' , 'label': 'node1_label'}, { 'id': 'node2_id' , 'label': 'node2_label'}]**
            'match_type' : 'external'
        },
       ...
     ]

It would be nice to save matches like this instead:

matches = [
        {    # External match where test_patient is the query and with results
            '_id' : 'match_1',
            'has_matches' : True,
            'data' : {
                'patient' : {
                    'id' : 'P0000079',
                    'contact' : {
                        'href' : 'mailto:[email protected]'
                    }
                }
            },
            **'results' : [ {
                   'searched_node' : {{ 'id': 'node1_id' , 'label': 'node1_label'}},
                   'patients' : [{'patient' : { 'patient_data' : 'test_stuff'}}, {'patient' : { 'patient_data2' : 'test_stuff2'}}]
             },
             {
                   'searched_node' : {{ 'id': 'node2_id' , 'label': 'node2_label'}},
                   'patients' : [{'patient' : { 'patient_data3' : 'test_stuff'}} ]
            },
          ],**# end of results
            'match_type' : 'external'
        },
       ...
     ]

New MME Service Requirements go into effect March 1.

Point n.7 :

For each database to which a MME service is connected by an API, the connected database’s disclaimers and terms should be posted on the MME service’s website and displayed with query results. Disclaimers can be found on GitHub (https://github.com/ga4gh/mme-apis). Disclaimers and terms may be returned with query results, which should supersede those found on the previously mentioned GitHub repository.

I'm still not sure what this means.

New point of the updated MME Service Requirements go into effect March 1:

Implement MME Metrics API (https://github.com/ga4gh/mme-apis/blob/master/metrics-api.md) and make metrics publicly available

I guess this will be done by creating a new endpoint. It should be relatively easy.

An arbitrary result threshold in the config file.

• Set it to 0 to return all hits
• Return only results above this threshold
• Number of max returned results (with highest score) is still decided by MAX_RESULTS param

• The current implementation of patientMatcher at Clinical Genomics returns the 5 patients most similar (with the highest patient score) to the patient used in the query.

• Any patient with a patient score above 0 might be returned, and the patient might be returned even if it has an extremely low score, for instance due to genotype not matching at all and phenotypes quite far in the HPO hierarchy

• Since it is better not to return any match then to send results with unspecific matches the SCORE_THRESHOLD should be fine-tuned. Action:

Return patients with a score compatible with at least a perfect gene (not variant) matching or an extremely similar phenotype to the query patient.

Current configs for patientMatcher @ CG:
MAX_GT_SCORE = 0.75
MAX_PHENO_SCORE = 0.25

Assuming that GT features are composed of 3 genes/variants, and assuming that there is gene matching for one of these features (1/4 of the max possible score) -->
0.75 / 3 (3 GT feature) / 4 --> 0.0625
with some phenotype similarity --> 0.08? 0.1?

0.0625 maybe to ensure that at least there is matching at the gene level (not variant).

Test on stage!

Send it here: https://github.com/ga4gh/mme-apis/tree/master/disclaimers

By assigning extra score if users provide at least 3 HPO terms.

Wait for feedback from the MME committee.

After ranking by similarity order them by most similar and return only those with highest similarity (or those sharing variants or genes with query patient, even if they have lower phenotype similarity)

For security reasons it's good to introduce an option to notify complete or partial info for MME matches by email.

I've changed the email body to follow the MME directives, see #65. Still it's good to have the option of NOT showing variants and phenotype terms of the matchings.

So add a config parameter that allows to show only contact info and cases ID in the notification emails.

Of course fix code that will become deprecated first

GT similarity is 0 when providing variant info

For instance if an external node sends a query patient x and there are 3 matches on the database (a, b, c) then the matching result is saved already. Create a method that allows you to ask the server: are there external matches for patient b? And the server returns external patient x

Fix also the tests accordingly

HPO terms: https://hpo.jax.org/app/download/ontology

Phenotype annotations: https://hpo.jax.org/app/download/annotation

Maybe it's interesting for final users. I am aware of at least another MME service which allows users to do that (RD Connect)

Make sure patientMatcher users are able to read email addresses from matches on other nodes

Try to insert 5000 patients (100 times those 50 benchmarking patients) and see how the server behaves.

But allow matching also with patients having genes described by entrez gene id and gene official symbol

So far ive assumed that server errors would also be returned a json but this is not universally valid because some node mighy return plain HTML, ad this causes a json parsing errror in the external matches handler. So do try to parse json but catch get html text as an option as well

This triggers an error in Scout matches view

Lowish prio

After a new software release today

I completely missed this: https://github.com/ga4gh/mme-apis/blob/master/heartbeat-api.md

Good thing it's easy to implement!

'_genotype' and '_phenotype' fields should be returned alongside 'patient' score

It's just not informative stuff and it 's just increasing the number of matching documents.

Here: https://github.com/ga4gh/mme-apis/wiki/Implementations

and configure nginx

Using the cli you should be able to see pmatcher version, not Flask version

Endpoints that should be reachable by external requests are:

1. server_url/metrics (GET method)

2. server_url/heartbeat (GET method)

3. server_url/match (POST method) *

Endpoints that should be reached from within the same server:

1. server_url/patient/add (POST method)

2. server_url/patient/delete/<patient_id> (DELETE method)

3. server_url/nodes (GET method)

4. server_url/match (POST method) *

5. server_url/match/external/<patient_id> (POST method)

6. server_url/matches (GET method)

• == endpoint which can be reached from the same server or by external request

By creating a new parameter in the settings. Default max number of matches returned = 10?

Should work from the cli and from DELETE requests

logging.warn should be replaced ---> The 'warn' function is deprecated, use 'warning' instead

Tested today after some time and it returns error (wrong json formatting?)

So that the module can be imported and used by other applications. Put the code that saves the matching objects to database outside the matching functions

MyGene2 is using ensembl ids for genes instead of approved symbols (like us) and when it tries to connect it looks like there are no matches even if our patients have the same gene

New MME Service Requirements go into effect March 1.

One of the updates is that:
If user receives a notification, the notification should include all information provided by the matching service or a direct link to additional available information.

This is relatively easy to fix. More info should be included in the match notification email

Both backend and frontend

Like a welcome page.
Things to add to this page:

• metrics, as in here: https://github.com/ga4gh/mme-apis/blob/master/metrics-api.md
• Disclaimer
• Link to howto (the github project repo)