PerturbRx: Learning Treatment-Conditioned Latent Transitions for Patient Drug Response Prediction
PerturbRx uses learned treatment-induced molecular shifts as the feature for predicting cancer drug response.
The method trains on control and treated single-cell populations matched by context, conditioning transitions on drug and dose. It then transfers that frozen transition predictor to pretreatment patient profiles, so it does not require post-treatment patient measurements. In TCGA and patient-derived xenograft benchmarks, the paper reports the strongest aggregate performance among the methods evaluated. ArXiv · AI/CL/LG's note
The method trains on control and treated single-cell populations matched by context, conditioning transitions on drug and dose. It then transfers that frozen transition predictor to pretreatment patient profiles, so it does not require post-treatment patient measurements. In TCGA and patient-derived xenograft benchmarks, the paper reports the strongest aggregate performance among the methods evaluated. ArXiv · AI/CL/LG's note
score 4