Nectin-4 PET for predicting enfortumab vedotin dose-response in urothelial carcinoma
Data files
Dec 11, 2025 version files 109 KB
-
README.md
5.50 KB
-
ROC_Curves_and_Response_prediction.zip
103.50 KB
Abstract
Optimizing dosing strategies is critical to balance effectiveness and toxicity, especially for drugs with narrow therapeutic windows such as antibody-drug conjugates (ADCs). This study evaluates whether positron emission tomography (PET) imaging targeting Nectin-4 can non-invasively quantify the real-time interaction of the ADC enfortumab vedotin (EV) with tumors in urothelial carcinoma (UC). Using the imaging agent [68Ga]AJ647, dynamic changes in interaction of EV with Necin-4 were measured across preclinical models and correlated with therapeutic responses. PET imaging identified dose-dependent variations in Nectin-4 engagement, with suboptimal EV doses resulting in incomplete Nectin-4 engagement and increased tumor growth. Crucially, PET measured target engagement predicted therapeutic outcomes more reliably than either drug dose or baseline target expression. By defining effective target engagement levels needed for optimal therapeutic outcomes, PET imaging provides a clear benchmark for dosing decisions, maximizing efficacy while potentially reducing exposure to higher, toxic doses and thereby enhancing patient safety.
Dataset DOI: 10.5061/dryad.612jm64jr
Overview
This README outlines the contents of files uploaded to Dryad for the study titled "Nectin-4 PET for predicting enfortumab vedotin dose-response in urothelial carcinoma." All required files are contained within the zip file “ROC_Curves_and_Response_prediction.zip".
1. Data File: 20240502_responder_prediction.xlsx
- Description: This file contains tumor volume (mm3) measurements for preclinical mouse models over time, structured to predict response using the modified RECIST criteria.
- File Structure:
- Each column (except the first) represents an individual mouse.
- The first column contains day numbers, and the first row identifies mouse IDs.
- Variables:
- Day (Column 1): Time points for tumor volume measurements. Unit for tumor volume is mm3.
- Mouse IDs (Row 1): Identifiers for individual mice.
2. Script: 20240429 Responders mRECIST prediction.ipynb
- Description: This Python script processes longitudinal tumor volume data and classifies responses according to modified RECIST-like criteria.
- Date Created: 2025-05-02
- Input File:
20240502_responder_prediction.xlsx - Note: The script automatically implements the criteria summarized from PMID: 26479923.
Requirements
- Python Version: 3.8+
- Packages:
- pandas
- openpyxl (for Excel I/O)
The script is designed to run smoothly without any additional requirements.
Output Categories and Criteria:
This Python script processes longitudinal tumor volume measurements in preclinical (mouse) models and automatically classifies each animal into one of four standard response categories using modified RECIST-like criteria. The criteria were summarized below:
ΔVol is calculated as:
ΔVol (%) = 100 × (V_day – V_day0) / V_day0 (relative to baseline, Day 0)
Only measurements from Day ≥ 10 onward are considered for average ΔVol response evaluation (to avoid early transient fluctuations).
For each time t, the average ΔVolt from t = 0 to t was calculated, with the ‘Best average ΔVol’ defined as the lowest average ΔVolt for t ≥ 10 days. ‘Best ΔVol’ is defined as minimum ΔVol ever achieved by a mice.
| Output Category | Criteria Applied in This Script |
|---|---|
| CR (Complete Response) | Best ΔVol ≤ −95% and Best average ΔVol ≤ −40% |
| PR (Partial Response) | Best ΔVol ≤ −50% and Best average ΔVol ≤ −20% |
| SD (Stable Disease) | Best ΔVol < +35% and Best average ΔVol < +30% |
| PD (Progressive Disease) | All other cases |
3. Text File: 20240429 Responders mRECIST prediction.txt
- Description: This text file contains the code corresponding to the
20240429 Responders mRECIST prediction.ipynbscript that performs longitudinal data processing and response classification. The contents and structure are similar to the Jupyter notebook, providing an alternative format for users.
4. Data File: 20240501_ROC_Curve.xlsx
- Description: This file serves as the input dataset for the ROC Curve Analysis.
- File Structure:
- Each row corresponds to a mouse, with columns representing different attributes as detailed below.
Variables:
| Column | Description | Type |
|---|---|---|
| M# | Mice number (identifier) | integer |
| dose | Pre-EV [68Ga]AJ647, unit is SUV | float |
| pre | Pre-intervention predictor score, unit is SUV | float |
| post | Post-EV [68Ga]AJ647, unit is SUV | float |
| predelpost | Difference or change score (post Ð pre), unit is SUV | float |
| engagement | Engagement metric (primary predictor), no unit, percentage | float |
| response | Response metric based on tumor volumes | integer |
5. Script: 20241210_ROC_Curves.ipynb
- Description: This script generates Receiver Operating Characteristic (ROC) curves and calculates optimal classification thresholds for predictive features based on tumor responses.
- Date: 2024-12-10
- Data File:
20240501_ROC_Curve.xlsx
Features Analyzed:
- pre: Pre-EV [68Ga]AJ647 SUV
- post: Post-EV [68Ga]AJ647 SUV
- predelpost: Change score (Post–Pre) [68Ga]AJ647 SUV
- engagement: -(Pre–Post)/Pre * 100
Requirements
- Python Version: 3.8+
- Packages:
- pandas
- numpy
- matplotlib
- scikit-learn
- plotly
- scipy
The script is designed to run smoothly without any additional requirements.
6. Text File: 20241210 ROC Curves.txt
- Description: This text file contains the code corresponding to the
20241210_ROC_Curves.ipynbscript for generating ROC curves and threshold calculations. The contents are structured similarly to the Jupyter notebook, providing an alternative format for users.