A foundation model can now simulate how a cell's entire genome will react to a new drug without ever touching a real petri dish.
Biologists usually spend months in the lab performing physical experiments to see how a genetic change affects a cell. This AI uses a Markov-chain sampling process to predict transcriptomic responses across the entire genome with high accuracy. It allows researchers to perform thousands of in-silico experiments in seconds to find the best drug candidates. This system handles complex multi-omics data to show the ripple effects of a single cellular perturbation. It moves drug discovery from a process of trial and error to a predictable digital simulation.
CellxPert: Inference-Time MCMC Steering of a Multi-Omics Single-Cell Foundation Model for In-Silico Perturbation
arXiv · 2605.00930
In this work, we introduce CellxPert, a scalable multimodal foundation model that unifies single-cell and spatial multi-omics within a common representation space. CellxPert jointly encodes transcriptomic (scRNA-seq), chromatin-accessibility (ATAC-seq), and surface-proteomic (CITE-seq) measurements, while directly incorporating MERFISH and imaging mass-cytometry data as 2D or 3D spatial-visual layers. CellxPert facilitates four key downstream tasks out of the box: (i) cell-type annotation across