Single-cell multimodal modeling with deep parametric inference

Abstract

The paired measurement of multiple modalities, known as the multimodal analysis, is an exciting frontier for connecting single-cell genomics with epitopes and functions. Mapping of transcriptomes in single-cells and the integration with cell phenotypes enable a better understanding of cellular states. However, assembling these paired omics into a unified representation of the cellular state remains challenging with the unique technical characteristics of each measurement. In this study, we built a deep parameter inference model (DPI) based on the properties of single-cell multimodal data. DPI is a complete single-cell multimodal omics analysis framework, which has built in multimodal data preprocessing, multimodal data integration, multimodal data reconstruction, reference and query, disturbance prediction and other analysis functions.