麻豆淫院

The rapid development of single-cell RNA sequencing technologies has made it possible to study the impact of external perturbations on gene expression at the level of individual cells.

However, in some cases, obtaining perturbed samples can be quite challenging, and the high cost associated with sequencing also limits the feasibility of large-scale experiments, requiring computational methods to predict single-cell gene expression perturbation responses.

For instance, leveraging existing data with perturbations induced by drugs to predict responses in new samples could provide valuable guidance for clinical diagnosis and treatment. Despite the existence of several methods, there is still room for further improvement in prediction accuracy.

To address these issues, the research team led by Shengquan Chen proposed SCREEN, a generative model based on masked variational autoencoder and optimal transport mapping. The work is in the journal Frontiers of Computer Science.

Comprehensive experiments on various datasets demonstrated that SCREEN significantly outperforms baseline methods in predicting single-cell perturbation responses.

In addition, the team showed the robustness of SCREEN to data noise, number of cell types, and cell type imbalance, indicating its broader applicability in various scenarios. They also demonstrated the ability of SCREEN to facilitate biological implications in downstream analysis, suggesting its great potential for single-cell perturbation analysis.