NMED-A143379A-Z

Code repository supporting Nature Medicine publication (Manuscript ID: NMED-A143379A-Z).

Overview

This repository contains the code and sample data used for the analyses presented in the accompanying manuscript. The codebase includes:

• Deep learning model architectures for quantitative image analysis
• Quantitative Continuous Scoring (QCS) feature extraction algorithms
• Statistical analysis pipelines for survival outcomes

Repository Structure

NMED-A143379A-Z/
├── README.md
├── data/
│   └── sampledata.csv          # Sample data for demonstration
└── notebooks/
    ├── 01_QCS_networks.ipynb   # Neural network architecture definitions
    ├── 02_QCS_readouts.ipynb   # QCS feature computation methods
    └── 03_feature_analysis.Rmd # Statistical analysis and visualization

Notebooks Description

01_QCS_networks.ipynb

PyTorch implementation of the deep learning architecture used for quantitative scoring.

02_QCS_readouts.ipynb

Algorithms for computing QCS features from segmented cell data:

• Quantile scores: Membrane, cytoplasm, and whole-cell staining intensity (SI) at quantile intervals (q5–q95)
• Percentage positive cells: Based on SI thresholds across membrane, cytoplasm, and whole-cell compartments
• Normalized Membrane Ratio (NMR): Membrane-to-cytoplasm SI ratios
• SI differences: Membrane minus cytoplasmic SI values
• Normalized SI values: Compartment-normalized intensity metrics
• QCS+ Cell Densities: Density of positive cells at various SI cutoffs
• Spatial proximity scores: Binary and continuous versions at multiple radii

03_feature_analysis.Rmd

R Markdown document for statistical analysis and figure generation:

• Kaplan-Meier survival curves for PFS and OS
• Cutoff optimization for normalized membrane OD quantiles
• Forest plots showing hazard ratios
• Ridge plots for interaction analysis

Requirements

Python

• Python 3.9+
• PyTorch
• torchvision
• NumPy
• SciPy
• pandas

R

• R 4.0+
• survival
• survminer
• ggplot2
• dplyr
• tidyr
• patchwork

Data Availability

Sample data is provided in data/sampledata.csv for demonstration purposes.

License

This project is licensed under the Apache 2.0 License - see the LICENSE file for details.

Contact

For questions regarding this repository, please contact Ansh Kapil.