How are associations scores calculated in the Open Targets Platform?

Following a number of user questions, we have just updated the documentation on target-disease association scoring.

Here is a quick summary of how our scoring works:

  • In the Open Targets Platform, an association is defined as a unique target-disease pair. Each association is supported by evidence.

  • Users can view scores for the evidence at different levels: at the data source level, at the data type level, and overall.

  • Scores at each level are calculated using a harmonic sum. This is then normalised to a score between 0 and 1.

For the data source and overall association scores, normalisation is a result of dividing the sum by the maximum theoretical harmonic sum, which approximates to 1.6. For the data type score, the maximum theoretical harmonic sum score is adjusted based on the number of data sources contributing to the data type.

  • The data source scores are used to calculate both the data type and overall association scores. In these calculations, the data source scores are weighted according to the data source weights we have defined.