Ground truth circularity in clinical trial validation – The CT415 dataset treats drugs already in trials as “positive,” but trial enrollment often reflects prior evidence or popularity (e.g., hydroxychloroquine), not true efficacy. This biases validation toward pipelines that rank well-known drugs highly, not necessarily correct ones. Did you control for trial start date or prior repurposing hype?
VeroE6 cell line confounding – VeroE6 cells lack type I interferon signaling, which alters antiviral drug sensitivity. Many “network drugs” might appear effective due to this deficiency rather than genuine host-target perturbation. Why was this cell line used for primary screening despite known artifacts for immunomodulatory drugs?