A novel colorimetric detection based on bifunctionalized gold nanoparticle combined with machine learning and deep learning models to identify microbial transglutaminase in foods.

Abstract

Microbial transglutaminase (mTG) is widely used in the food industry to enhance the appearance and texture of meat and fish products, as well as the smoothness and richness of dairy products. However, the undisclosed excessive addition of mTG contributes to various health issues, including celiac disease with intestinal leakage, anemia, osteoporosis, dermatitis, and other parenteral symptoms. In this study, we developed a novel method combining gold nanoparticles (AuNPs), machine learning, and deep learning to study mTG activity in both aqueous solutions and diverse processed foods. Our results demonstrate that this colorimetric method, based on bifunctionalized AuNPs, exhibits sufficient sensitivity to detect pure mTG down to 0.01U and spans a detection range from 0.01U to 1U. Based on the colorimetric changes of gold nanoparticles, we constructed a dataset of 648 mTG concentration-absorbance data points from 6 different food types. We employed machine learning algorithms, including Decision Tree (DT), Random Forest (RF), and Multilayer Perceptron (MLP), to predict mTG concentration based on the colorimetric signal in various foods. Notably, the MLP model achieved a high prediction accuracy of 0.96. Blind tests on six types of supermarket-purchased meat, seafood, and dairy products showed predictions consistent with expected mTG levels. This study establishes an efficient strategy for the identification and prediction of mTG activity in a wide range of food products.