PySingleCellNet: classify scRNAseq data in Python

SingleCellNet predicts the ‘cell type’ of query scRNA-seq data by Random forest multi-class classification. See Tan 2019 for more details. It was originally written in R. PySCN is the Python version which includes functionality to aid in the analysis of engineered cell populations (i.e. cells derived via directed differentiation of pluripotent stem cells or via direct conversion).

User guide

• Getting started
  • Prerequisites
  • Installation
• Prepare training data
  • Summary
  • Training data
  • Load packages and raw data
  • Quality control
  • Normalize and reduce dimentionality
  • Embed data and cluster
  • Cluster annotation
• Train and classify
  • Data
  • Load packages
  • Load data
  • Limit to genes shared in both data sets
  • Train classifier
  • Classify held out data and assess
  • Classify query data
  • Cell type composition
  • Rank-based classifier
• Explore results
  • Data
  • Load packages
  • Load training data and query data
  • Limit to genes shared in both data sets
  • Split up training data, and then train classifier
  • Assess performance on heldout data
  • Now analyze the query data
  • Empirical categorization of classification calls
  • Gene set enrichment analysis
• GSEA & annData
  • Data
  • Gene signatures
• Steps to GSEA
  • Load packages
  • Load gene signatures
  • Load data
  • Norm/HVG/PCA
  • kNN/Cluster/UMAP
  • Annotate these ‘coarse’ clusters
  • Analysis of coarse clusters
  • Analysis of cell types
• Training data
  • Mouse embryo data
  • Mouse training data from Tabula senis 10X.
• References