Artificial Intelligence and MALDI-TOF Spectrometry Improve Bacteria and Virus Detection
Researchers from Greece, Estonia, and Belgium have improved bacteria and virus classification by combining artificial intelligence with MALDI-TOF mass spectrometry.
A research team from universities in Greece, Estonia, and Belgium has improved the classification of bacteria and viruses by combining artificial intelligence with MALDI-TOF mass spectrometry. However, tests with external datasets revealed challenges in the cross-laboratory transfer of AI models.
AI and MALDI-TOF Integration
Researchers from Greece, Estonia, and Belgium combined artificial intelligence with MALDI-TOF mass spectrometry to improve the classification of bacteria and viruses. The study aimed to enhance the potential of this method, which enables rapid identification in clinical microbiology using protein fingerprints.
Model Tests and Accuracy Rates
The researchers evaluated eight machine learning and two deep learning models using mass spectra obtained from seven bacterial species and five viral agents. On the internal dataset, many of the models achieved a one hundred percent accuracy rate in distinguishing between bacteria and viruses.
External Dataset and Encountered Challenges
The top three strongest models were tested with an external dataset from the Robert Koch Institute. Differences in cross-laboratory calibration, sample preparation, and inactivation reduced performance, while species-level identification proved more challenging.
Future Studies and Expectations
The researchers acknowledged that the small internal dataset increases the risk of overfitting. Larger and more diverse datasets from multiple laboratories will be needed for the widespread real-world adoption of AI-supported classification.