Deep generative models, such as variational autoencoders (VAE), have gained increasing attention in computational biology due to their ability to capture complex data manifolds which subsequently can be used to achieve better performance in downstream tasks, such as cancer type prediction or subtyping of cancer. However, these models are difficult to train due to the large number of hyperparameters that need to be tuned. To get a better understanding of the importance of the different hyperparameters, we examined six different VAE models when trained on TCGA transcriptomics data and evaluated on the downstream task of cluster agreement with cancer subtypes. We studied the effect of the latent space dimensionality, learning rate, optimizer and initialization on the quality of subsequent clustering of the TCGA samples. We found β-TCVAE and DIP-VAE to have a sensitive to hyperparameters selection. Based on these experiments, we derived recommendations for selecting the different hyperparameters settings. In addition, we examined whether the learned latent spaces capture biologically relevant information. Hereto, we correlated the different representations with various data characteristics such as age, days to metastasis, immune infiltration, and mutation signatures. We found that for all models the latent factors, in general, do not uniquely correlate with one of the data characteristics even for models specifically designed for disentanglement

Motivation: Single-cell multi-omics assays simultaneously measure different molecular features from the same cell. A key question is how to benefit from the complementary data available and perform cross-modal clustering of cells. Results: We propose Single-Cell Multi-omics Clustering (scMoC), an approach to identify cell clusters from data with co-measurements of scRNA-seq and scATAC-seq from the same cell. We overcome the high sparsity of the scATAC-seq data by using an imputation strategy that exploits the less-sparse scRNA-seq data available from the same cell. Subsequently, scMoC identifies clusters of cells by merging clusterings derived from both data domains individually. We tested scMoC on datasets generated using different protocols with variable data sparsity levels. We show that, due to its imputation scheme, scMoC 1) is able to generate informative scATAC-seq data due to its RNA guided imputation strategy, and 2) results in integrated clusters based on both RNA and ATAC information that are biologically meaningful either from the RNA or from the ATAC perspective.

Motivation Single-cell multi-omics assays simultaneously measure different molecular features from the same cell. A key question is how to benefit from the complementary data available and perform cross-modal clustering of cells. Results We propose Single-Cell Multi-omics Clustering (scMoC), an approach to identify cell clusters from data with co-measurements of scRNA-seq and scATAC-seq from the same cell. We overcome the high sparsity of the scATAC-seq data by using an imputation strategy that exploits the less-sparse scRNA-seq data available from the same cell. Subsequently, scMoC identifies clusters of cells by merging clusterings derived from both data domains individually. We tested scMoC on datasets generated using different protocols with variable data sparsity levels. We show that scMoC 1) is able to generate informative scATAC-seq data due to its RNA-guided imputation strategy, and 2) results in integrated clusters based on both RNA and ATAC information that are biologically meaningful either from the RNA or from the ATAC perspective. Availability The code is freely available at: https://github.com/meltager/scmoc. Supplementary information Supplementary data is available online.