Single-cell preprocessing with omicverse

What problem does it solves? Preparing raw single-cell RNA-seq data for analysis involves numerous critical steps like quality control, normalization, and dimensionality reduction, which are often computationally intensive and require careful parameter tuning. This Skill automates and accelerates these processes.

Core Features & Use Cases

• Comprehensive Quality Control: Perform QC filtering, including doublet detection (Scrublet), to ensure high-quality data.
• Normalization & HVG Detection: Normalize counts, log-transform, and identify highly variable genes (HVGs) for feature selection.
• Dimensionality Reduction & Embedding: Compute PCA, build neighborhood graphs, and generate UMAP/MDE embeddings for visualization.
• GPU Acceleration: Support for CPU, CPU-GPU mixed, and pure GPU (RAPIDS) processing for faster workflows.
• Use Case: Take raw 10x Genomics PBMC3k data, apply QC filters, normalize and select HVGs, then generate a UMAP embedding, leveraging GPU acceleration for faster processing.

Quick Start

Preprocess my PBMC3k single-cell data, including QC with Scrublet, normalize with shiftlog|pearson, and generate a UMAP embedding.