ISG Screen Analysis

Analysis code for the live-cell imaging (Incucyte) ISG screen described in:

The ISG Atlas: A Loss-Of-Function Analysis Characterizes Antiviral Properties of Interferon Stimulated Genes — Krey et al., Nature Communications, 2026.

---

Overview

This repository contains the analysis pipeline for a genome-wide ISG (Interferon-Stimulated Gene) screen performed using the Incucyte live-cell imaging system. The screen tested ~300 ISGs against 7 viruses to identify candidates with antiviral activity.

The fluorescence readout — Green Integrated Intensity / Red Integrated Intensity (GII/RII) — serves as a proxy for viral replication: a higher GII/RII ratio indicates greater reporter expression and therefore stronger ISG-mediated restriction.

Viruses screened:

Code	Virus
V1	HSV-1
V2	MeV
V3	RVFV
V4	VACV
V5	YFV
V6	VSV
V7	SFV
V8	SARS-CoV-2

---

Repository structure

├── main_analysis.ipynb   # Main analysis notebook (run this)
├── requirements.txt      # Python dependencies
├── data/
│   ├── fluorescence/     # Incucyte GII/RII export files (virus-infected plates)
│   └── viability/        # Incucyte GII export files (mock-infected control plates)
└── src/
    ├── read_data.py       # I/O helpers for Incucyte .txt exports
    ├── process_signal.py  # Signal denoising (median + Savitzky-Golay smoothing)
    ├── fit_data.py        # Curve fitting (sigmoid, exponential, linear)
    └── virus2mock.py      # Mapping of virus plates to matched mock controls

Data file naming convention

V<virus>_S<set>_R<replicate>.txt

• V — virus ID (0 = mock/uninfected control)
• S — set number (1–4, each set corresponds to one 96-well plate of ~75 ISGs)
• R — biological replicate number

---

Analysis pipeline

The notebook main_analysis.ipynb runs the following steps:

1. Viability screen — exponential fits to GII on mock-infected plates identify knockouts that cause cellular toxicity. Toxic KOs are excluded from downstream analysis.
2. Fluorescence data loading — GII/RII time-series are loaded for all virus–ISG–replicate combinations.
3. Signal cleaning — early-timepoint noise is removed and the signal is smoothed with a median + Savitzky-Golay filter pipeline.
4. Sigmoid fitting — a logistic growth model is fit to each time-series to extract parameters: K (amplitude), τ (induction time), β (growth rate), and c (baseline).
5. Normalisation — fitted parameters are normalised to the mean of 5 non-targeting controls (NTCs) per plate.
6. Merging — normalised fluorescence and viability parameters are merged into a single output DataFrame.

---

Setup and usage

Requirements

• Python ≥ 3.9

Installation

1. Install Python or install conda.

2. Clone this repository:

   git clone <repo-url>
   cd isg_analysis

3. Create a virtual environment:

   python -m venv isg_screen

4. Activate the environment:

   • Unix/macOS: source isg_screen/bin/activate
   • Windows: .\isg_screen\Scripts\Activate

5. Install dependencies:

   pip install -r requirements.txt

6. Launch the notebook:

   jupyter notebook main_analysis.ipynb

More information on virtual environments is available in the Python documentation.