Capturing data…
Capturing data…
MON, 12 OCT · 83 ITEMS
"Transcriptional perturbation effects remain structured beyond the leading global trend" — bioRxiv (Genomics) · Science & Medicine
This is a finding from a new preprint benchmarking models that predict how cells respond to genetic or chemical perturbations. It shows that simple linear models can capture the dominant global changes in gene expression (the 'leading trend,' analogous to the first principal component), but the finer, perturbation-specific structures beyond that trend remain and are missed by those approaches, meaning more sophisticated models are needed to fully capture perturbation effects.
The claim comes directly from a bioRxiv preprint (also appearing on arXiv), which is a primary source reporting the authors' own analysis of model performance on perturbation response datasets. As a preprint, it has not yet undergone peer review, but the finding is a direct empirical result from the study's benchmarks.
Genuinely new: the preprint appears to have been posted in 2026 (arXiv ID 2604.27646) and was highlighted from a recent bioRxiv upload.