All data is harmonised, analysed and merged using a combination of programming languages (Python v3, R v3.3, Scala v2.12) and computational libraries and frameworks (e.g. Apache Spark v2.4.5, GCTA v1.93) and stored in Parquet format when possible. The resulting 2TB of inferences are loaded into ClickHouse v20.1.4.14 and ElasticSearch v5.6.16 (Supplemental Figure S1). These two services are configured, loaded and optimised in GCP. More importantly, they are released publicly to users interested in creating their own instances of the Genetics Portal. The API is also available at https://genetics-api.opentargets.io and implemented using GraphQL v2.0.0, Play Framework v2.7.3, Slick framework v3.3.2 and Sangria v2.0.0. The web application is available at https://genetics.opentargets.org and utilises React v16.8 as well as a number of javascript libraries with emphasis on D3.js v5.5 for custom interactive visualisations. The large amount of data also introduces challenges for some user interface components. To build the locus plot (Supplemental Figure S2) for example, we use a canvas element as it is not possible to accomplish the visualisation with regular DOM nodes. To ensure global access, Open Targets Genetics is deployed
