Prototyping a Flexible Antibiotic-Loaded Colorimetric Biosensor and Assessing Colorimetric Detection to External Stimuli via Machine Learning

Our research work focuses on developing a colorimetric biosensor integrated into smart textiles, emphasizing the application of machine learning (ML) techniques to enhance detection capabilities for environmental changes like change in temperature and pH level, and bacterial presence, particularly Escherichia coli. The biosensor uses polydiacetylenes (PDAs), specifically 10,12-Pentacosadiynoic acid (PCDA), within an electro spun core-shell nanofiber structure as the colorimetric indicator. The shell comprises PCDA, Polyurethane (PU), and Polyethylene Oxide (PEO), ensuring structural integrity and rapid colorimetric responsiveness to stimuli. The core that is made of PEO, encapsulates penicillin nanoparticles within a porous Poly (lactic-co-glycolic acid) (PLGA) structure, facilitating targeted antibacterial drug release upon detection of bacterial presence.

A significant part of this project is to optimize the PCDA concentration. Through systematic experimentation, we aim to find out the ideal concentration of PCDA, a high-cost material, that not only maximizes the colorimetric response of the biosensor but also minimizes its response time to environmental changes, including fluctuations in temperature and pH levels, and bacterial detection. This optimization is essential for enhancing the biosensor's performance while ensuring cost-effectiveness in its industrial application.

To enable real-time monitoring, machine learning algorithms, specifically Convolutional Neural Networks (CNNs), have been employed to analyze the colorimetric data generated by the biosensor. The initial phase involves data collection through experiments that capture colorimetric responses under known controlled conditions. High-resolution images of the sensor's color changes have been quantified using image processing algorithms, providing a robust dataset for ML model training. Following data collection, preprocessing steps have been implemented to refine the dataset, including noise reduction and feature extraction. The selected features have been used to train CNN models capable of recognizing patterns related to external stimuli and predicting pathogen presence. Based on this data, the machine learning program will analyze and detect the results of unknown samples. The performance of these machine learning models will be evaluated based on their accuracy in interpreting sensor data, with a specific emphasis on sensitivity testing against established benchmarks.

The integrated biosensor will also undergo performance testing in various textile materials, assessing its effectiveness and response time in real-world scenarios for environmental monitoring and bacterial detection. Importantly, this biosensor not only detects E. coli bacteria but also actively kills them through the controlled release of penicillin, thereby addressing both identification and mitigation of bacterial threats. This dual approach—focusing on optimizing PCDA concentration along with the response time and implementing advanced machine learning techniques—provides a comprehensive solution for active monitoring in healthcare, food safety, and environmental contexts.

By merging the unique properties of PDAs with sophisticated machine learning methodologies, this project represents a significant advancement in biosensing technology. The anticipated outcomes include the development of intelligent biosensors capable of dynamically responding to environmental changes while providing reliable pathogen detection and intervention. Integrating the ML program will also eliminate the need for human involvement in interpreting results from manual colorimetric biosensors, enabling automated data analysis and immediate result generation through machine. This research addresses critical challenges in public health and food safety, paving the way for innovative applications in smart textiles that enhance real-time monitoring capabilities.