NNMF-for-scRNAseq

Non negative matrix factorization for Oligo lineage scRNAseq data

Analysis Pipeline for scRNA-seq Data Using NMF

This repository contains an R script (script.R) for analyzing single-cell RNA sequencing (scRNA-seq) data using Non-negative Matrix Factorization (NMF) with Seurat and RcppML libraries. Below is a step-by-step guide on how to reproduce the analysis:

Steps to Reproduce

1. Load Required Libraries

   • Seurat: library(Seurat)
   • dplyr: library(dplyr)
   • RcppML: library(RcppML)

2. Set Seed for Reproducibility

   • set.seed(200)

3. Load Seurat Object

   • Replace /data/nasser/Manuscript/processedobject/ODC35_woClus8_subclust3_res0.15_NK with your own path.
   • ol <- readRDS("/data/nasser/Manuscript/processedobject/ODC35_woClus8_subclust3_res0.15_NK")

4. Set Cell Type Identities

   • Idents(ol) <- "CellType"

5. Subset the Data

   • Choose specific cell types for analysis (iODC, iOPC, iPPC_0, iPPC_1, iPPC_2).
   • Uncomment and modify if subsetting by iCEP is required.
   • pd <- subset(ol, idents = c("iODC", "iOPC", "iPPC_0", "iPPC_1", "iPPC_2"))

6. Optional: Visualize Using DimPlot

   • Uncomment DimPlot(pd) to visualize the subset data.

7. Clean the Cluster (Optional)

   • Clean UMAP coordinates to focus on specific areas (umap1 > -2 & umap2 > 1).

8. Extract Expression Matrix

   • Extract the RNA expression matrix from the subsetted Seurat object.
   • expression_matrix <- LayerData(object = pd, assay = "RNA", layer = "data")
   • Remove rows with NA or null values and those with all zero values.

9. Set Number of Clusters

   • Determine the number of clusters based on unique cell types.
   • num_clusters <- length(unique(pd$CellType))

10. Perform NMF

    • Apply NMF to the expression matrix.
    • nmf_result <- nmf(expression_matrix, k = num_clusters, tol = 1e-4, maxit = 500)

11. Extract Basis (W) and Coefficient (H) Matrices

    • Retrieve the basis (W) and coefficient (H) matrices from the NMF result.
    • W <- nmf_result$w
    • H <- nmf_result$h

12. Identify Most Influential Genes

    • Determine top influential genes for each cluster.
    • Store results in influential_genes list.

13. Convert Results to Data Frame

    • Convert the list of influential genes to a tidy data frame (influential_genes_df).

14. Example Visualization

    • Generate feature plots for each cluster using influential genes.

Notes

• Adjust paths (readRDS) and parameters (subset, nmf) based on your specific data and analysis requirements.
• Ensure all necessary libraries (Seurat, dplyr, RcppML) are installed and loaded before running the script.