scICER Workflow: Example Workflows and Manuscript Figure Scripts for Integrated Single-Cell Analyses

Introduction

scICER is an R package for evaluating cluster consistency in single-cell RNA-seq data. It reimplements the scICE methodology to assess the stability of cluster labels across stochastic runs. Fully integrated with Seurat, scICER works on embeddings corrected by Harmony or scVI enabling consistent clustering analysis in complex, multi-sample datasets.

Data sources, preprocessing, and conversion

The repository includes source tables, dataset preparation notes, and preprocessing scripts for the benchmark analyses, pancreas integration examples, the large pancreas atlas from GSE221156, and the runtime comparison panels.

The corresponding materials are organized under data/, including:

• data/README.md
• data/preprocessing/
• data/tables/

Codes examples to run scICER

The codes/ directory contains example R scripts demonstrating the use of scICER in several common analysis settings.

The workflow examples include:

• codes/run_scICER.R
• codes/harmony_integration_scICER.R
• codes/scvi_integration_scICER.R

These scripts are written as reusable examples and use generic inputs rather than manuscript-specific output paths.

Manuscript figure scripts

The codes/figures/ directory contains the R scripts used to reproduce the manuscript figures. Scripts are organized by figure, and each .R file corresponds to one figure block.

The figure scripts are accompanied by:

• codes/figures/README.md
• source tables in data/tables/