About

UMAP-based visualization and clustering pipelines for single-cell RNA-seq data (like the Seurat pbmc3k tutorial) are highly sensitive to their input parameters. Small changes in resolution, number of PCs, or neighbor counts can produce drastically different plots, making it difficult for users to objectively identify the most appropriate configuration for their dataset.

Our tool addresses that problem by using an evolutionary algorithm to systematically search for an optimal set of pipeline parameters for the given dataset, guided by the Calinski-Harabasz index. Upload your dataset, run the optimization, and explore the resulting visualization.

Associated Paper (unpublished): geneticUMAP: Objective and Robust Parameter Optimization for scRNA-Seq Data Visualization
Authors: Lisa-Maria Winter, Dominik Dreiheller, Hanna Wagner, Julian Schwab, Johann M. Kraus, and Hans A. Kestler
Software Architect & Developer: Hanna Wagner | Developers: Dominik Dreiheller, Lisa-Maria Winter

Quick Start

Click on the button above to upload your single-cell dataset (10x or h5 file format) and start a new optimization job. You can then inspect the optimized UMAP visualization and review how the CH-index evolved across generations for each job by opening the tab that will be created for your job after successful data upload.

Upload Note

For data upload, this tool supports the two native Seurat import functions, Read10x() and Read10x_h5(), which are specifically optimized for the output generated by the 10x Genomics Cell Ranger pipeline.

All other single-cell RNA-seq data formats must be converted into either the 10x matrix format or HDF5 (.h5) format prior to upload.

The 10x matrix format requires three separate files:

a .tsv file containing cell barcodes (unique identifiers for individual cells),
a .tsv file listing gene annotations, and
a .mtx file containing the count matrix.

Note: Successful data import requires that the dimensions of the count matrix exactly match the entries in both, the gene and barcode file. Gene and barcode files should not contain any additional columns.

Security Notice

Please be aware that any data entered into this web application is uploaded and stored on our server for 30 days. Anyone with a valid job id may access and download the corresponding data, since there is no authorization mechanism.

If you are handling sensitive or private information, please use the R package geneticumap implementing the underlying optimization algorithm to ensure your data remains within your own secure environment.