Within the booming field of single-cell omics analysis, spatial transcriptomics is becoming more and more popular. Being awarded method of the year 2020 by Nature, it has a bright future ahead. The method will put us a step closer to understanding the spatial organization of cells in tissues, and how the organization influences cell function. At this point, no spatially aware transcriptome-wide method has reached single-cell level yet, complicating the data analysis. The first step in this project will be to overcome this hurdle. Different types of single-cell omics data can shed a complimentary light on a tissue, but current methods can't go beyond the integration of two data modalities. Variational autoencoder based methods, creating joint latent spaces, seem suitable for this problem, but their black-box nature is a limitation. In this project, novel machine learning model frameworks that can integrate multiple data modalities (single-cell RNAseq, CITE-Seq, and spatial transcriptomics) and result in biologically interpretable latent spaces, will be built. The framework will be flexible and easily adapted to new technological developments in the spatial transcriptomics field. The variables in the interpretable latent space can be linked to specific cellular aspects. This will allow the study of correlations between those aspects, and answering more complex biological research questions.