Contains code used to generate the figures and the analysis included in the Starlng manuscript.
The code relies on the usage of the Starlng package.
Methods that were investigated in the manuscript: Starlng, hdWGCNA, Hotspot, SCENIC, PAGA + CellRank.
Structure of the repository:
benchmarking: test the runtime and memory usage of Starlng against immune-based synthetic datasets (10k-100k cells);data: preprocessing steps for the real datasets and generation of the synthetic datasets using Sergio;comparison_against_ground_truth: analysis of the similarity between the method results and the ground truth.comparison_with_other_methods: analysis of the output of each clustering method based on the 5 metrics described in the manuscript.moran_i_impact: exploratory analysis of the distribution of Moran's I values and its relationship to other cohesion metrics such as pseudotime iqr and median umap distance.robustness_analysis: analysis of the robustness of Starlng against subsampling, noise adding and seed changes.starlng_run: scripts to run Starlng on real datasets and generate the figures included in the manuscript.