Validation depth
Evidence breadth across identification strategies, from computational inference to randomised clinical evidence.
Computational eQTL, LLM-inferred 2 edges
Pre-clinical Knockout (KO)
Observational Cohort, case-control
Clinical RCT, Mendelian randomisation
LLM-inferred 2
Causal edges (2)
Upstream drivers pointing at white matter preservation as the downstream effect.
- drug:CXCL10 inhibition → white matter preservation LLM-inferred low
- drug:CXCR3 blockade → white matter preservation LLM-inferred low