🧬 scRNAseq-Figure-Replication

BF550 Final Project

Project Overview

This project replicates key figures from the study:

Rubenstein, A. B., Smith, G. R., Raue, U., Begue, G., Minchev, K., Ruf-Zamojski, F., Nair, V. D., Wang, X., Zhou, L., Zaslavsky, E., Trappe, T. A., Trappe, S., & Sealfon, S. C. (2020). Single-cell transcriptional profiles in human skeletal muscle. Scientific reports, 10(1), 229. https://doi.org/10.1038/s41598-019-57110-6

The original paper investigates gene signatures across mononuclear and multinucleated skeletal muscle cell types using single-cell RNA sequencing (scRNA-seq).

Unlike bulk RNA-seq, single-cell analysis enables:

• Identification of cell-type–specific transcriptional signatures
• Detection of heterogeneous gene expression patterns
• Separation of changes in cell-type proportions from gene expression differences

---

Objective

The goal of this project was to computationally replicate:

• Supplementary Figure S9
• Figure 3a

using:

• Hierarchical clustering
• PCA (Principal Component Analysis)
• Louvain clustering
• tSNE visualization
• Marker gene dot plots
• Heatmaps

All analyses were performed in Python using Scanpy.

---

Data Source

Raw count matrices were downloaded from GEO (Gene Expression Omnibus).

• Four skeletal muscle scRNA-seq count files were merged
• Final dataset contained 2,876 cells, matching the published study

---

Computational Workflow

Data Processing

Counts Matrix Generation

• Raw count .csv.gz files were loaded using pandas
• Gene names were set as the dataframe index
• Datasets were merged using an outer join
• Missing values were removed
• Data was transposed to cell × gene format
• Converted into an AnnData object

Final AnnData object:

• 2,876 cell observations
• Gene expression matrix
• Metadata annotations

---

Preprocessing (Scanpy)

The following Scanpy preprocessing steps were applied:

• normalize_total() → library size normalization
• log1p() → log transformation
• highly_variable_genes() → feature selection
• pca(n_comps=10) → dimensionality reduction
• neighbors() → graph construction using 10 PCs

---

Unsupervised Clustering

• Louvain clustering (resolution = 0.6)
• Clusters initially unlabeled
• tSNE computed for visualization

sc.tl.louvain()
sc.tl.tsne()

---

Figure Replication

---

A. tSNE Clustering (Figure 2A Equivalent)

Unlabeled clusters were visualized using tSNE.

Cluster identities were determined using:

• Marker gene expression
• Dot plots
• Hierarchical clustering via dendrogram

Clusters were manually annotated by mapping:

cluster ID → biological cell type

Final labeled tSNE shows:

• Endothelial subtypes
• Satellite cells
• Pericytes
• NK cells
• T + B cells
• Myeloid cells
• Smooth muscle cells
• FAP subtypes

---

B. Marker Gene Dot Plot

To identify cell types:

• Top marker genes per cluster (from Supplementary Table 1) were used

• Scanpy dotplot() visualized:

  • Expression level (color intensity)
  • Fraction of expressing cells (dot size)

• Dendrogram computed using PCA correlations between clusters

This step guided manual cluster annotation.

---

C. Hierarchical Clustering

Hierarchical clustering was computed using:

sc.tl.dendrogram()

Clustering was based on:

• Correlation of PCA components
• Cluster-level aggregation

This matches the hierarchical grouping described in the paper.

---

D. Heatmap (Figure 3a Replication)

A curated marker gene dictionary was constructed using:

• Top-ranked differentially expressed genes
• Supplementary Table 1 from the paper

Heatmap generated using:

sc.pl.heatmap()

The resulting heatmap:

• Displays marker expression across annotated clusters
• Recapitulates published cell-type–specific transcriptional signatures
• Reflects hierarchical relationships between cell types

---

Biological Interpretation

This replication demonstrates:

• Clear transcriptional separation of skeletal muscle cell subtypes
• Distinct endothelial subpopulations
• Immune cell presence (NK, T, B, Myeloid)
• Fibro-adipogenic progenitor (FAP) heterogeneity
• Satellite cell signature specificity

The results confirm the original study’s findings regarding cell-type–specific gene expression patterns.

---

Technologies Used

• Python
• Scanpy
• Pandas
• NumPy
• scikit-learn
• Matplotlib
• Seaborn
• SciPy

---

Key Learning Outcomes

This project demonstrates:

✔ Single-cell RNA-seq preprocessing

✔ AnnData object construction

✔ Feature selection via highly variable genes

✔ Dimensionality reduction (PCA, tSNE)

✔ Graph-based clustering (Louvain)

✔ Marker-based cell type annotation

✔ Hierarchical clustering

✔ Figure replication from published literature