scFPCDE-paper-code

This repository contains the analysis code used to reproduce results from the manuscript scFPC-DE: Differential Expression Analysis Along Single-Cell Trajectories via Functional Principal Component Analysis.

Repository structure

• data_analysis/ contains the HB6 B-cell pseudotime reconstruction and the numbered real-data analysis scripts. Run the numbered scripts in order.
• simulation/ contains the simulation-study code extracted from the supplementary material.

HB6 input modes

Install the current scFPCDE package before running the numbered scripts. The HB6 preprocessed data are stored in this repository, keeping the R package focused on its reusable methods and simulation example. The loader chooses between two reproducible input modes:

1. Local full objects. If both data/scPure2_HB6_UMAP3D.rds and data/cds_HB6.rds exist, 0_load_data.R loads them and script 1 reproduces the cell-quality, pseudotime, variable-gene, and detection-filtering steps.
2. Repository preprocessed data. If those files are absent, 0_load_data.R loads data/scFPCDE_hb6.rda. This contains the exact two already-subsetted paper trajectories, with aligned raw counts, uncentered logcounts, pseudotime, and B-cell labels. Script 1 starts from those inputs and reruns scFPCDE and PseudotimeDE.

Run from the repository root:

source("data_analysis/0_load_data.R")
source("data_analysis/1_trajs_analysis_workflow.R")

Then run scripts 2 through 12 in numerical order. Script 1 can also be sourced directly; it will invoke the loader when needed.

To choose a mode explicitly before starting R:

SCFPCDE_PAPER_DATA_MODE=preprocessed R
SCFPCDE_PAPER_DATA_MODE=local R

auto is the default. Preprocessed mode begins after trajectory reconstruction and cell/gene selection, so bcell_hb6_pseudotime.R is needed only when rebuilding the full Monocle trajectory from the original data. All required R packages must still be installed before running the analyses.