Dataset for "Machine Learning–Guided Design of Mechanoadaptive Bioglues for Multi-Tissue Trauma and First-Aid Applications"

This repository contains the raw data, machine learning models, and experimental results supporting the study published in Nature Biomedical Engineerin (or relevant journal based on final publication). The study introduces TuneGlues, a suite of polyurethane-based bioglues with customizable mechanical properties designed for multi-tissue trauma. By leveraging machine learning (ML), we established task-oriented relationships between bioglue formulations and tissue characteristics, enabling precise optimization for specific targets (lung, intestine, skin, bone, etc.). Additionally, this dataset supports the development of a data-based first-aid device capable of rapidly delivering optimized TuneGlues.
Data Contents
The dataset is organized into the following categories:
1.Source Data for Figures: Raw numerical data underlying all main text and supplementary figures (e.g., mechanical testing, adhesion strength, burst pressure).
Machine Learning Dataset: - Experimental dataset of 96 TuneGlue formulations with measured mechanical properties (modulus, strength, deformation).
- ML-predicted database containing 100,000 virtual formulations.
- Model architecture and training parameters (NN, RFR, XGBoost, etc.).
Biological Evaluation: - In vitro cytocompatibility data.
- In vivo animal study records (porcine and caprine models) including surgical logs, healing observations, and histological analysis data.
4.Device Specifications: Design files and control parameters for the portable data-based first-aid device.
5.Supplementary Information: Complete supplementary tables (S1-S4), figures (S1-S33), and video descriptions (V1-V10) as referenced in the manuscript.
Code Availability
The code necessary to reproduce the machine learning experimental findings and model training can be found on GitHub:
🔗 [https://github.com/ChaoyuCao/TuneGlue.git](https://github.com/ChaoyuCao/TuneGlue.git)

Usage Instructions
Mechanical Data: Files are provided in `.csv` or `.xlsx` formats. Column headers describe the formulation components (SP1, SP2, HP1, HP2, SC, HC) and resulting properties.ML Models: Python scripts (PyTorch, scikit-learn) are included for model training and prediction. See the GitHub repository for detailed dependencies.Statistical Analysis: Statistical methods (ANOVA, Student's t-test) are documented in the "Methods" section of the associated paper.