Disease Progression Trajectory Analysis

Disease Progression Trajectory Analysis

When to Use This Skill

Use this skill when you have longitudinal patient omics data and want to:

• ✅ Reconstruct disease progression trajectories from time-series data
• ✅ Order samples by disease stage (pseudotime) with irregular sampling
• ✅ Identify biomarkers changing along disease trajectory
• ✅ Stratify patients as fast vs. slow progressors
• ✅ Predict clinical outcomes from trajectory position
• ✅ Validate computational staging against clinical measures

Required data:

• Minimum 10 patients with 3+ timepoints each
• Omics data: RNA-seq, proteomics, metabolomics, or clinical biomarkers
• Metadata: Patient IDs, timepoints (days/months/years), optional outcomes

Primary method: TimeAx multiple trajectory alignment (handles irregular sampling)

Feature identification: Polynomial regression (linear/quadratic/cubic) per the TimeAx paper (Frishberg et al., Nat Commun 2023), with FDR-corrected Q-value filtering. Captures both monotonic and non-monotonic dynamics.

Alternative methods: Linear Mixed Models (regular sampling), Hidden Markov Models (discrete stages)

Installation

R ≥ 4.0 with the TimeAx package (primary trajectory method):

# Install TimeAx from GitHub
install.packages("remotes")
remotes::install_github("amitfrish/TimeAx")

# Required for plotting
install.packages(c("ggplot2", "ggprism"))

# Required for demo dataset (GSE128959 batch correction)
BiocManager::install("sva")

Python ≥ 3.9 for the workflow wrapper and analysis pipeline:

# Core analysis packages
pip install numpy pandas scipy scikit-learn statsmodels lifelines

# Visualization packages
pip install seaborn matplotlib

# PDF report generation (optional)
pip install reportlab

# Optional
pip install hmmlearn        # Hidden Markov Models alternative

For Linear Mixed Models alternative: R packages lme4, lmerTest

License compliance: All packages use permissive licenses (MIT, BSD, Apache 2.0) - commercial AI agent use permitted.