Cell Dynamics Analysis

Overview

This skill answers: How do cells change over time, and what are their fate decisions?

Two complementary approaches:

1. RNA Velocity: Infers directionality from spliced/unspliced ratios
2. Pseudotime/Trajectory: Orders cells along developmental paths

Decision Tree: Which Approach?

Q: Does your data have spliced/unspliced information?
│
├─ YES (velocyto/kallisto processed)
│   │
│   └─ RNA Velocity Analysis
│       ├─ scVelo (deterministic) - Fast, good default
│       ├─ scVelo (dynamical) - More accurate, slower
│       └─ VeloVI - Deep learning, handles noise
│       │
│       └─ Then: CellRank for fate probability
│
└─ NO (standard scRNA-seq/spatial)
    │
    └─ Pseudotime Analysis
        ├─ Palantir - Multi-lineage, fate probabilities
        ├─ DPT (Diffusion Pseudotime) - Classic, fast
        └─ PAGA - Trajectory + clustering

RNA Velocity Workflow

Prerequisites

Data must contain:

• adata.layers['spliced']: Spliced counts
• adata.layers['unspliced']: Unspliced counts

Check with: 'spliced' in adata.layers and 'unspliced' in adata.layers

Step 1: Preprocessing for Velocity

Velocity-specific preprocessing:
1. Filter genes by spliced/unspliced detection
2. Normalize spliced and unspliced separately
3. Compute moments (first/second order)

Step 2: Compute Velocity

Use analyze_velocity_data tool with method selection:

Mode	Speed	Accuracy	When to Use
deterministic	Fast	Good	Initial exploration
stochastic	Moderate	Better	Publication, noisy data
dynamical	Slow	Best	Complex dynamics, time recovery

Step 3: Visualize Velocity

Use visualize_data with plot_type="velocity":

• subtype="stream": Velocity streamlines on embedding
• subtype="phase": Phase portraits for specific genes
• subtype="proportions": Spliced/unspliced ratios
• subtype="heatmap": Velocity across pseudotime

Step 4: Fate Analysis with CellRank

After velocity computation:

1. Use analyze_trajectory_data with method="cellrank"
2. Identify initial/terminal states
3. Compute fate probabilities
4. Identify lineage driver genes

Pseudotime Workflow (No Velocity Data)

Method Selection

Method	Strengths	When to Use
Palantir	Multi-lineage, fate probs	Complex differentiation
DPT	Fast, robust	Simple trajectories
PAGA	Topology discovery	Unknown structure

Step 1: Define Root Cell

Critical for pseudotime:

Options for defining root:
1. Known marker gene expression (e.g., stem cell markers)
2. Specific cluster known to be progenitor
3. User-specified cell barcode

Step 2: Compute Trajectory

Use analyze_trajectory_data tool with:

• method: "palantir", "dpt", or "cellrank"
• root_cell or root_cluster: Starting point

Step 3: Visualize Trajectory

Use visualize_data with plot_type="trajectory":

• subtype="pseudotime": Cells colored by pseudotime
• subtype="fate_map": Fate probability heatmap
• subtype="gene_trends": Expression along trajectory
• subtype="circular": Circular trajectory plot

Biological Interpretation

What Velocity Tells You

• Direction of differentiation: Where cells are heading
• Dynamic genes: Genes actively changing
• State transitions: Intermediate cell states
• Terminal states: End points of differentiation

What Pseudotime Tells You

• Ordering of cells: Relative developmental time
• Branching points: Fate decision locations
• Lineage relationships: Which states lead to which
• Driver genes: Genes driving transitions

Spatial Context

Combining Velocity with Space

Questions to explore:

• Do velocity vectors align with spatial gradients?
• Are terminal states localized to specific regions?
• How does spatial organization relate to developmental time?

Visualization Tips

• Overlay velocity arrows on spatial coordinates
• Color spatial plots by pseudotime
• Identify spatial zones of active differentiation

Common Analysis Scenarios

Tumor Heterogeneity

Focus on:
- Cancer stem cell identification
- Differentiation hierarchies within tumor
- EMT transitions

Development

Focus on:
- Lineage specification
- Progenitor identification
- Temporal gene programs

Regeneration/Wound Healing

Focus on:
- Activation states
- Re-differentiation paths
- Spatial gradients of repair

Troubleshooting

Issue	Cause	Solution
No velocity vectors	Low unspliced counts	Check data quality, try VeloVI
Random directions	Steady state cells	Focus on dynamic subpopulations
Wrong root assignment	Biological uncertainty	Try multiple roots, validate with markers
Disconnected trajectory	Batch effects	Integrate samples, increase neighbors

Output Summary

Successful dynamics analysis provides:

• Pseudotime values: Ordering of cells
• Velocity vectors: Direction of change
• Fate probabilities: Likelihood of terminal states
• Driver genes: Key regulators of transitions
• Spatial mapping: Where dynamics occur in tissue