SARS-CoV-2 Infection Dynamics - Single-Cell RNA-seq Analysis
Reference: Ravindra et al., PLOS Biology (2021) - Single-cell longitudinal analysis of SARS-CoV-2 infection in human airway epithelium
https://doi.org/10.1371/journal.pbio.3001143
Data: GEO GSE166766 (10x mtx/tsv)
https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE166766
🎯 Project Objective
Reproduce neighborhood clustering, cell-type annotation, and trajectory/pseudotime analysis of SARS-CoV-2 infection in HBECs (human bronchial epithelial cells). Focus on mock, 1 dpi, 2 dpi, and 3 dpi, targeting reproduction of figures similar to 1G(i-iii), 3A/B, 4A/B.
Special emphasis:
- Comprehensive pseudotime analysis to capture cell state transitions over infection.
- Clear annotation of ionocytes as a rare but biologically relevant population.
🔬 Biological Overview
- Eight epithelial cell types identified: ciliated, basal, club, BC/club (basal→club), goblet, neuroendocrine, ionocytes, tuft.
- Ciliated cells are the primary targets of early infection. Infection spreads to basal, club, BC/club, and ionocytes at later stages.
- ACE2 expression indicates cell-type susceptibility but is a poor per-cell predictor of infection.
- ENO2 marks neuroendocrine/metabolic state changes and cellular stress, offering complementary insights into infected vs. bystander cells.
✅ Analysis Workflow (Modular Approach Recommended)
1. Data Loading
def load_adata(path, var_names='gene_symbols'):
"""Load 10x matrix into AnnData object with error handling."""
import scanpy as sc
import os
if not os.path.exists(path):
raise FileNotFoundError(f"Data path {path} not found.")
adata = sc.read_10x_mtx(path, var_names=var_names, cache=True)
return adata
- Load each condition (mock, 1 dpi, 2 dpi, 3 dpi) separately.
- Avoid system-specific download commands; use Python requests or GEOquery to fetch data.
2. Quality Control
- Filter cells with <200 genes and genes present in <3 cells.
- Remove cells with >15% mitochondrial reads.
- Use modular QC functions with inline comments for clarity.
3. Normalization & Feature Selection
- Normalize total counts → CPM → log1p.
- Identify highly variable genes (HVGs, n=2000).
- Encapsulate preprocessing in reusable functions:
def preprocess_adata(adata):
import scanpy as sc
sc.pp.normalize_total(adata, target_sum=1e4)
sc.pp.log1p(adata)
sc.pp.highly_variable_genes(adata, n_top_genes=2000, flavor='seurat_v3')
return adata[:, adata.var['highly_variable']]
4. Dimensionality Reduction & Batch Correction
- PCA → BB-kNN (preserves neighborhood across batches).
- UMAP & PHATE embeddings for visualization and pseudotime.
5. Clustering & Annotation
- Louvain clustering → manual annotation using canonical markers:
- FOXJ1 → ciliated
- KRT5 → basal
- SCGB1A1 → club
- MUC5AC → goblet
- PAX6 / ENO2 → neuroendocrine
- POU2F3 → tuft
- FOXI1 → ionocytes
- Modular functions for annotation allow future reuse.
6. Infection Flagging
- Cells with ≥10 viral transcripts flagged as infected.
- ACE2 vs ENO2 expression analyzed per cluster for mechanistic insights.
7. Pseudotime / Trajectory Analysis
- Compute diffusion pseudotime (DPT) using PHATE embeddings.
- Overlay pseudotime on UMAP/PHATE to interpret infection progression.
- Optional: RNA velocity with scVelo if spliced/unspliced counts are available.
def compute_pseudotime(adata, n_dcs=10):
import scanpy as sc
sc.tl.diffmap(adata)
sc.tl.dpt(adata, n_dcs=n_dcs)
return adata
🔍 Key Observations
- Cell types at each infection stage:
- Mock: All eight types, including ionocytes, present; no viral signal.
- 1 dpi: Infection primarily in ciliated cells.
- 2-3 dpi: Spread to basal, club, BC/club, and ionocytes.
- Correlation with infection:
- Tropism is cell type-driven via ACE2 and TMPRSS2.
- Pseudotime analysis reveals transitions in cell states and temporal infection patterns.
- ACE2 vs ENO2:
- ACE2: Entry receptor; indicates cell susceptibility.
- ENO2: Metabolic/neuroendocrine stress marker; highlights state changes in infected/bystander cells.
- ACE2 enrichment (3 dpi):
- Highest in ciliated cells, overlaps with infected populations.
- Suggests primary entry sites and persistent susceptibility.
📦 Deliverables
- README.md - this improved overview
- notebook_clustering.ipynb - modular, annotated preprocessing, clustering, and annotation
- notebook_pseudotime.ipynb - detailed trajectory/pseudotime analysis
- figures/ - UMAP/PHATE, heatmaps, violin plots (reproducing 1G, 3A/B, 4A/B)
- results/ - per-cluster ACE2 expression & infected cell counts
📚 References