A Machine Learning-Based Diagnostic Utility for Predicting Ertapenem Resistance in Klebsiella Pneumoniae Using MALDI-TOF MS Spectra

Klebsiella pneumoniae is a gram-negative bacterium that causes over 600,000 deaths annually worldwide. Carbapenems, a powerful class of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\beta$</tex>-lactam antibiotics, are used to treat Klebsiella pneumoniae, as a lastresort when other antibiotics fail. However, when bacteria become resistant to carbapenems, treatment options become very limited. Ertapenem, a commonly used member of the carbapenems family, is widely applied in clinical settings, making early detection of ertapenem resistance crucial for preventing the spread of resistant strains and improving patient outcomes. Traditional methods such as Polymerase Chain Reaction, Whole-Genome Sequencing, and Gene Expression Analysis remain limited because they require long processing times, rely on complex and high-cost equipment, and may not reliably detect new/uncommon resistance mechanisms. Newer approaches such as MALDI-TOF MS (Matrix-Assisted Laser Desorption/Ionization Time-of-Flight Mass Spectrometry) offer rapid bacterial identification, but determining whether isolates are resistant or susceptible to ertapenem still requires additional time-consuming manual processes. In this study, we integrated several machine learning (ML) algorithms: Logistic Regression, K-Nearest Neighbors (KNN), Multi-Layer Perceptron (MLP), and Random Forest, with MALDI-TOF MS spectra to predict ertapenem resistance with high accuracy. The results were promising: Logistic Regression, MLP, and Random Forest achieved testing accuracies of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$86 \%, 84 \%$</tex>, and 78 %. This approach has strong potential to improve the diagnostic capability of MALDI-TOF MS by enabling rapid and accurate detection of Klebsiella pneumoniae resistance and supporting timely, targeted decision-making. Unlike previous research, which has been more generic in scope, this study is specifically focused on ertapenem (an easy-to-administer, cost-effective carbapenem) and employs lightweight ML models that can run on standard laboratory computers, making the method more suitable for real-world clinical applications.