Brain Single-Cell Analysis Repository

Method Benchmark: scGPT vs. CellTypist

This repository includes a method benchmark comparing scGPT and CellTypist for the task of brain cell. The benchmark compares both tools in terms of cell-type prediction performance.

You can find the code and detailed results in the provided scripts for both scGPT and CellTypist. Refer to the classification/method_benchmark files for more information.

This dataset contains a benchmark comparision of methods os cell type annotation alongside with single cell data analysis.

Reference Datasets:

• hiPSC Data: The hiPSC data used as a reference is sourced from the study published in Nature Neuroscience (PubMed ID: 37451260). The dataset can be accessed and read using the Cell Ranger file format with scanpy’s read_10x function.
• Full Fetal Reference: The fetal reference data used in this repository is derived from the study published in Science (DOI: 10.1126/science.adf1226).

Packages Used:

• Python:
  • scanpy (for reading and preprocessing single-cell RNA data)
  • celltypist (for cell-type annotation using pre-trained models)
  • anndata (for handling single-cell data formats)
  • scGPT (for cell-type annotation)
• R:
  • SingleR (for cell-type annotation in R)