MultiVI multimodal generative model capable of integrating multiome, scRNA-seq and scATAC-seq data. After training, it can be used for many common downstream tasks, and also for imputation of a missing modality.
The advantages of multiVI are: (i) Comprehensive in capabilities. Able to perform DE gene, DA region analysis. (ii) Scalable to very large datasets (>1 million cells). (iii) Once trained with sufficient multimodal data, able to accurately input missing modalities.

Important: MultiVI requires the datasets to use shared features. scATAC-seq datasets need to be processed to use a shared set of peaks. This features dataset must be included in the inputs

Example use case: MultiVI can be used for many common downstream tasks, and also for imputation of a missing modality

Limitations: Effectively requires a GPU for fast inference.