A systematic control bias inflates mean-baseline performance in scRNA-seq perturbation benchmarks, and the proposed DEG-weighted metrics with an all-perturbed-cells reference yield null mean-baseline performance while improving GEARS when used as a training loss.
Deep learning-based predictions of gene perturbation effects do not yet outperform simple linear methods
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
q-bio.GN 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Diversity by Design: Addressing Mode Collapse Improves scRNA-seq Perturbation Modeling on Well-Calibrated Metrics
A systematic control bias inflates mean-baseline performance in scRNA-seq perturbation benchmarks, and the proposed DEG-weighted metrics with an all-perturbed-cells reference yield null mean-baseline performance while improving GEARS when used as a training loss.