RATarget — Target Identification & Scoring

Interpretation notice. This produces a ranked hypothesis list, not a validated prediction. Weights are user-chosen and have not been calibrated against any outcome dataset — no ground truth for "good target" exists that could be used honestly here. Treat the score as a way to make your prioritisation criteria explicit and reproducible, and always inspect the component columns rather than the composite alone.
Safety is off by default. Open Targets liability records mix genuine adverse findings with a target's intended pharmacology — KCNJ11, for instance, lists "decreased blood glucose". Section 3 lets you decide which terms count before the component contributes anything.

1 · Disease

Search disease / phenotype

Or enter an EFO / MONDO ID directly Include evidence from disease subtypes Targets to retrieve Protein-coding targets only

2 · Evidence weights

Weights are normalised to sum to 1, so only their ratios matter. Re-scoring is instant — no API call is made when you move a slider.

3 · Safety

Safety measure

Remaining limitations, unfixed: severity is unweighted, so arrhythmia and hypoglycaemia each count 1. And a target with no records scores 0.5 ("unknown"), so the measure still partly rewards being unstudied.

4 · Other evidence datatypes

Run a query first — these are read from whatever datatypes the API actually returns.

Datatypes Open Targets returned that the fixed components above do not already use. All default to zero weight.

5 · Prioritisation factors

Run a query first — the available factors are read from whatever the API returns.

From Open Targets' target prioritisation panel, scored −1 (unfavourable) to +1 (favourable). All default to zero weight. Normalised as (v+1)/2 before weighting.

6 · Ranked targets

Data diagnostics — raw values returned by the API