Fig. 4From: Scalable transcriptomics analysis with Dask: applications in data science and machine learningRuntime comparison between the Dask and SPE frameworks in A the preprocessing and B the full loading of scRNA-seq data. Datasets were subsampled for different dimensions. In A, both Dask Distributed configurations (Threads and Processes) partially load the data, processing dataset partitions; in B, the entire dataset is loaded. Asterisks represent instances when programs ran out of memory. n is the number of rows and f the number of features. The file sizes shown in GB correspond to uncompressed tabular filesBack to article page