Fig. 1

A workflow diagram of the DEGAS framework. A The workflow for a typical experiment with DEGAS. Note that DEGAS is not meant to replace the abundant packages available to load, preprocess, select features, cluster, and visualize scRNA-seq data. It is rather meant to augment these packages to assign disease associations to cells. B The scRNA-seq and patient expression data are preprocessed into expression matrices. Next, Bootstrap aggregated DenseNet DEGAS models are trained using both single-cell and patient disease attributes using a multitask learning neural network that learns latent representation reducing the differences between patients and single cells at the final hidden layer using maximum mean discrepancy (MMD). C The output layer of this model can be used to simultaneously infer disease attribute impressions in single cells and cellular composition impressions in patients