InClust+: the multimodal version of inClust for multimodal data integration, imputation, and cross modal generation

Lifei Wang, Rui Nie, Yankai Cai, Anqi Wang, Hanwen Zhang, Jiang Zhang, Jun Cai
bioRxiv (2023)


With the development of single-cell technology, many cell traits (e.g. gene expression, chromatin accessibility, DNA methylation) can be measured. Furthermore, the multi-omic profiling technology could jointly measure two or more traits in a single cell simultaneously. In order to process the various data accumulated rapidly, computational methods for multimodal data integration are needed. Previously, we developed inClust, a flexible all-in deep generative framework for transcriptome data. Here, we extend the applicability of inClust into the realm of multimodal data by adding two mask modules: an input-mask module in front of the encoder and an output-mask module behind the decoder. We call this augmented model inClust+, and apply it to various multimodal data. InClust+ was first used to integrate scRNA and MERFISH data from similar cell populations and to impute MERFISH data based on scRNA data. Then, inClust+ is shown to have the capability to integrate a multimodal data contain scRNA and scATAC or two multimodal CITE datasets with batch effect. Finally, inClust+ is used to integrate a monomodal scRNA dataset and two multimodal CITE datasets, and generate the missing modality of surface protein in monomodal scRNA data. In the above examples, the performance of inClust+ is better than or comparable to the most recent tools to the corresponding task, which prove inClust+ is a suitable framework for handling multimodal data. Meanwhile, the successful implementation of mask in inClust+ means that it can be applied to other deep learning methods with similar encoder-decoder architecture to broaden the application scope of these models.