single-cell-clustering-and-batch-correction-with-omicverse

You can perform single-cell clustering and batch correction on AnnData by applying preprocessing, feature selection, and neighbourhood graph construction, followed by methods like Leiden, Louvain, scICE, or GMM clustering alongside Harmony, scVI, BBKNN, or Combat integration.

Comparing clustering methods across batches involves applying Leiden, Louvain, scICE, and GMM strategies to identify robust groupings, then evaluating clustering stability and ARI improvements against reference labels after applying batch correction.

Yes, you can use batch integration methods like Harmony, scVI, BBKNN, and Combat to harmonize datasets. The workflow supports comparing these strategies to evaluate ARI improvements and visualize embeddings against reference labels.

Single-cell clustering with OmicVerse requires reproducible workflows including data preservation by saving to adata.raw. It also performs scalable preprocessing and defensive checks for dependencies before applying clustering and batch correction.

You need batch correction for single-cell clustering when combining multiple datasets to remove technical variation. Applying Harmony, scVI, BBKNN, or Combat ensures robust cell states are derived by harmonizing data before downstream analysis and evaluation.

Yes, alongside clustering and batch correction, you can apply advanced modeling options like topic modeling (LDA) and cNMF to uncover underlying cell programs, which supports benchmarking and explainable clustering within your analysis.