AI-assisted electrochemical detection of bacteria siderophores in clinical samples

The rapid detection of bacteria remains a significant challenge in clinical contexts, with consequences for both patients and the medical system. While the electrochemical sensors deliver fast results, the matrix effect in real samples and the overlaps of the electrochemical signals are still of major concern and cannot be overcome in the absence of biorecognition elements and samples pretreatment protocols. The implementation of portable sensors in this case has promising potential, but their use is still limited to laboratory settings. An innovative approach is to develop an AI-assisted electrochemical sensor that employs machine learning algorithms to classify the target signal and overcome the interferences from real samples with complex biological matrices. The electrochemical behavior of three bacterial markers for Pseudomonas aeruginosa , Klebsiella pneumoniae , and Escherichia coli was analyzed in human serum to estimate the impact of any potential interference, whether from the environment or between components. Chemometrics have been used to evaluate the influence that the markers have on each other's analytical signals in the case of their simultaneous detection. A Decision Tree Classifier algorithm has been applied to enhance the sensitivity and improve the accuracy of the classification in cases of complex samples with promising results on real clinical samples. • Simultaneous detection of bacteria markers was achieved via electrochemical analysis. • Chemometrics analysis identified the mutual interference effect of the molecules. • Machine learning algorithms were applied to enhance the accuracy of classification. • The training and testing data sets were done in commercial human serum.