Based on the discussion under #3282 and #3283, the parser to generate disease target evidence from Genetics Portal data needs to be updated with the following things:
[x] From QTL integration exclude splice QTLs.
[x] When selecting a representative QTL for an evidence, select based on the most significant P-value instead of the highest effect.
Background
The rationale behind this update is that the interpretation of sQTLs is very complicated and cannot derive direction of effect. Also by chosing biggest effect, the sample size of the QTL study skews towards poorly powered studies. By picking the lowest p-value, we are going for the highest power + effect combination. All these changes are expected to improve the predictive power of direction of effect assessments.
Acceptance tests
Validate newly created disease target evidence
Make sure no splice QTLs are included and ensure the selection was done based on p-value
Based on the discussion under #3282 and #3283, the parser to generate disease target evidence from Genetics Portal data needs to be updated with the following things:
Background
The rationale behind this update is that the interpretation of sQTLs is very complicated and cannot derive direction of effect. Also by chosing biggest effect, the sample size of the QTL study skews towards poorly powered studies. By picking the lowest p-value, we are going for the highest power + effect combination. All these changes are expected to improve the predictive power of direction of effect assessments.
Acceptance tests