HAPI-Pheno Server

A HAPI FHIR implementation for working with the GA4GH Phenopacket FHIR Implementation Guide.

The code is based on the hapi-fhir-jpaserver-starter project, and many classes were copied. The server was extended with functionality to work withthe Phenopacket FHIR IG.

Setup: IG

Install SUSHI, SUSHI Unshortens ShortHand Inputs as described on the SUSHI webpage. Install jekyll as described on the webpage.

Clone the core-ig repository.

Run _updatePublisher.sh to install the FHIR publisher tool from the core-ig repo.

Run _genonce.sh from the core-ig. This creates a folder output with the published IG. Note the path to this directory, which we will set in the application.properties file.

Running the server

Clone the hapi-fhir-jpaserver-starter project. There are many ways of starting the server, but we will use the following (shown as <output> below).

mvn -Djetty.port=8888 jetty:run

The server will then be accessible at http://localhost:8888/ and eg. http://localhost:8888/fhir/metadata.

Loading LOINC Code/Value sets

It is absolutely necessary to load LOINC data before running the Phenopackets example. To do so, download the latest Loinc_2.72.zip archive from the LOINC website. (requires an account)

Then download the hapi-fhir CLI tool from the HAPI Releases page. If you downloaded source, build it: mvn package, but the major releases usually have a distribution. There's even one for just the cli here: https://github.com/hapifhir/hapi-fhir/releases/download/v6.0.1/hapi-fhir-6.0.1-cli.zip Then run the following command (note that the -t command points to the running server endpoint).

java -jar hapi-fhir-cli.jar upload-terminology \
  -d Loinc_2.72.zip \
  -v r4 \
  -t http://localhost:8888/fhir \
  -u http://loinc.org

This will take roughly 30 minutes to complete. The command returns quickly, the server will be busy for a while.

With LOINC 2.73, I got the following error: a.uhn.fhir.rest.server.exceptions.UnprocessableEntityException: HTTP 422 Unprocessable Entity: Could not find the following mandatory files in input: [AccessoryFiles/MultiAxialHierarchy/MultiAxialHierarchy.csv] Fix this by navigating to the Archive of past LOINC releases, and get 2.72 until this is fixed in HAPI-FHIR.

Note that we added the following to the Application.java class

/**
 * ERROR c.u.f.jpa.term.BaseTermReadSvcImpl [BaseTermReadSvcImpl.java:1796] Failed to pre-expand ValueSet: maxClauseCount is set to 1024
 * org.apache.lucene.search.BooleanQuery$TooManyClauses: maxClauseCount is set to 1024
 */
int MAX_CLAUSE_COUNT = 10_000; // to avoid problems as above
BooleanQuery.setMaxClauseCount(MAX_CLAUSE_COUNT);

It is also necessary to uncomment the #hibernate.search.enabled and following lines in the application.yaml file (already done here).

Examining the structure definitions

http://localhost:8888/fhir/StructureDefinition

Retrieving a Phenopacket and coding as GA4GH native phenopacket

See the accompanying project