Map single-cell datasets into reference atlases using scArches.
npx skills add https://github.com/Ketomihine/my_skills --skill scarches
Please help me install this Agent Skill. Skill: scarches Source: https://github.com/Ketomihine/my_skills/tree/main/scarches Command: npx skills add https://github.com/Ketomihine/my_skills --skill scarches
This Skill enables researchers to integrate and map new single-cell datasets into a curated reference atlas, enabling cross-study analyses, batch-aware embeddings, and consistent downstream interpretation.
Prepare reference and query AnnData objects, install scArches and dependencies, select a model (SCVI, TRVAE, scANVI, totalVI, mvTCR, or scPoli), train on the reference dataset, and perform surgery/mapping on the query dataset to obtain integrated embeddings and transferred annotations.
High-intent search queries and answers about installing and using this skill.
Mapping new single-cell datasets to a reference atlas involves training a model like SCVI or scANVI on the reference AnnData, then performing atlas surgery to integrate the query dataset and transfer annotations.
Atlas surgery in single-cell analysis is the process of surgically adapting a trained reference model to align query datasets, enabling cross-study integration, batch-aware embeddings, and consistent downstream interpretation.
Yes, scArches supports multi-omics integration across scRNA-seq, scATAC-seq, CITE-seq, and spatial chromatin data through specialized models like totalVI and mvTCR for cross-study atlas mapping.
Yes, you need the scanpy ecosystem and scvi-tools to perform batch-aware embedding and batch correction on scRNA-seq data, as these frameworks provide the foundational models for scArches mapping workflows.
Available models for cross-study integration and label transfer include SCVI, TRVAE, scANVI, totalVI, mvTCR, and scPoli, which train on reference datasets and map query data for harmonized embeddings.
The best way to detect novel cell states across multiple studies is to perform robust batch correction and joint clustering using atlas mapping, which harmonizes embeddings and enables comparative analysis across conditions.
LeonChaoXTrain probabilistic generative models on AnnData counts for single-cell omics analysis.
jackspaceModel single-cell omics data with scvi-tools for batch-corrected analysis.
Victory-HugoRun scvi-tools deep generative models on single-cell RNA sequencing data.
omicverseConvert bulk RNA-seq data into synthetic single-cell profiles with Bulk2Single.
