Score values from Disease->Target vs. Target->Disease

I was looking for the association score of MAPT to neurodegenerative disease. When I search for “Neurodegenerative Disease” (Open Targets Platform) and then filter to MAPT, the score comes up as 0.86. However, when I start from MAPT and filter for neurodegenerative disease (Open Targets Platform) the score comes up as 0.01.
Please could you explain the difference?
Thank you!

Hello @Heather! :wave:

Welcome to the Open Targets Community!

The difference in the scores presented in the UI is due to how we handle direct and indirect evidence. By this, I mean how we handle evidence for the specific disease/phenotype (direct) and evidence for a child term of the disease/phenotype (indirect).

When you search for “neurodegenerative disease”, we show you the list of associated targets and the overall score and data type scores are based on both direct evidence (e.g. text mining for MAPT and neurodegenerative disease) and indirect evidence (e.g. genetic association evidence for MAPT and Parkinson’s disease, a child term of neurodegenerative disease).

However, when you search for MAPT, we show you the list of associated diseases and phenotypes and the overall score and data type scores are based on direct evidence only (e.g. text mining for MAPT and neurodegenerative disease).

Despite the different scores depending on your search term, the evidence page for MAPT and neurodegenerative disease contains both direct and indirect evidence.

Please let me know if this has answered your question.

Thank you! :slight_smile:

Thanks so much for your reply @ahercules !
I do have a few follow-on questions that I hope you can also help with. Firstly, when I use the GraphQL API (GraphQL Playground) the results I get show “score” and dont specify whether the score is direct or indirect (I assume it is an overall score though). Performing a disease search I get this:
And when I use the target search I see:
But I dont see how I can know which score is direct and which is not - it is possible to know?

The other question is regarding the type of evidence - is it possible to see the scores coming from different evidence sources using the GraphQL API?

Thanks once again,


Hi @Heather!

No, unfortunately it’s not possible to see direct and indirect scores using the API. Similar to what you will see in the UI, if you start your query with the target endpoint (second screenshot), your results will include associations only based on direct evidence and so the score will only be based on direct evidence. If you start your query with the disease endpoint (first screenshot), your results will include associations based on both direct and indirect evidence and the score will be based on both types of evidence.

If you are keen on getting separate direct and indirect association scores, you can use our data downloads and sample Python and R scripts to access and query specific datasets for direct and indirect associations.

Alternatively, you can also use our Google BigQuery instance to explore the direct and indirect datasets for a given target-disease association and export the results in various formats. See below for example queries for EGFR and lung cancer using both the direct and indirect datasets.

If you run both of the above queries, you will see how the chembl and europepmc scores are different depending on whether you are using direct or indirect datasets.

In regards to your second question, if you are using the GraphQL API, you can query for the per data source scores using the following query:

query targetAssociations {

Sample query for EGFR


Andrew :slight_smile: