OCOO-T : A Simple and Scalable Virtual Cell Model for Transcriptional Perturbation Response Prediction
Researchers have introduced OCOO-T, a new virtual cell model designed for predicting transcriptional responses to various perturbations. This model utilizes a simplified Transformer architecture and a flow-matching approach, treating response prediction as a continuous-time denoising process. OCOO-T integrates perturbation details and cell-specific information through adaptive layer normalization and in-context tokens, demonstrating state-of-the-art performance on multiple benchmarks while maintaining scalability for long transcriptional profiles. AI
IMPACT Simplifies complex biological simulations, potentially accelerating drug discovery and gene regulatory network analysis.