: A neural network framework that predicts transcriptional responses to both single- and two-gene perturbations.

: A model that extends perturbation studies from static snapshots to dynamic cellular trajectories, allowing for the simulation of disease progression or development.

: A causally inspired graph neural network that identifies which combinations of perturbations are needed to reverse a disease phenotype. Software & Frameworks

In the context of biology and machine learning, a "perturbation" typically refers to an experimental intervention—such as a genetic knockout or chemical treatment—that alters a cell's state to study its response.

Several recent papers and frameworks focus on predicting these responses using machine learning: Key Research Papers (2024–2026)

: A machine learning architecture designed to predict cellular responses to perturbations across diverse biological contexts.

: A meta-learning framework that translates existing perturbation atlases to predict responses in new biological contexts using only a few "seed" perturbations.