Analysing and Visualising Single Cell Omits Data

Single cell multimodal omics data are characterized by sparsity, noise, and high dimension. Incorporating information across modalities is challenging. We developed a non-negative matrix factorization based algorithm to identify latent factors that can facilitate improved cell-type clustering and visualizations for multimodal single cell omics count data. We then extend this algorithm to larger datasets and for different distributions of data in different modalities.