scArches

About

Provides expert assistance with scArches, a specialized deep learning library for single-cell analysis and data integration. This skill enables users to efficiently process scATAC-seq and scRNA-seq data, perform batch correction, and map query datasets to established reference atlases using advanced 'model surgery' techniques. It offers comprehensive implementation guidance for architectures like scVI, SCANVI, and TRVAE, helping researchers navigate complex workflows such as multi-modal integration, cell-type label transfer, and spatial chromatin analysis while ensuring optimal hyperparameter selection and model convergence.

Key Features

• Automated cell-type label transfer using latent representation and KNN classification
• Model surgery workflows for adapting pre-trained references to new query data
• 0 GitHub stars
• Technical guidance on loss functions, regularization, and preprocessing pipelines
• Implementation patterns for scVI, SCANVI, TRVAE, and totalVI architectures
• Multi-modal integration for transcriptomics, chromatin accessibility, and TCR data

Use Cases

• Training deep generative models for multi-omic single-cell data analysis
• Mapping experimental query data to reference atlases for rapid cell-type identification
• Integrating disparate single-cell datasets while mitigating batch effects