Poster on Single-Cell data Analysis via Perturbation Estimation
Date:
RNA velocity provides the directional information of cell-state dynamics and transitions from single-cell RNA-sequencing data. To reveal cellular transitions and perturb this process in silico, models capable of unveiling the causal effects of complex gene regulatory interactions are required. Here, we present SCAPE, a deep learning-based model that predicts the transcription dynamics under in silico genetic perturbations. SCAPE is uniquely able to predict in silico post-perturbation RNA velocities, which can be directly used to predict cell-fate diversions induced by the genetic perturbation of certain genes. We applied SCAPE to predict cell-fate diversions after in silico genetic perturbations in neurogenesis, gastrulation, and pancreatic endocrinogenesis. We validated the efficacy and accuracy of SCAPE in recapitulating well-characterized cell-state transitions induced by gene perturbations. Overall, SCAPE will facilitate the biomedical study of cell-fate decisions and cell-state transitions.