All data in the Open Targets Platform is publicly available. However some datasets require ad hoc pipelines to perform mapping or post-processing steps required to align them to the current data model. A JSON validator ensures all evidence respects the schema. All annotation, evidence and ontologies are then processed by the data pipeline to reconstruct entities, clean and score evidence and associations, generate search indexes and calculate disease-disease similarities. The extract-transform-load pipeline is written in Python v3 and the results loaded in ElasticSearch v7.6. The REST API is currently available at https://platform-api.opentargets.io/ and implemented using Python v2 and Flask framework v1.1. The web application is written using Angular 1.7 as well as a number of libraries such as D3.js v.3.5 for interactive visualisations. To ensure global access, the infrastructure is deployed across three different regions. Each regional deployment is the same and globally balanced. All our services are configured, loaded and optimised in Google Cloud Platform and our code is open source and accessible in the repositories listed in Supplementary Table S3.
