Duque Mathilde, Emeraud Cécile, Bonnin Rémy A, Giai-Gianetto Quentin, Dortet Laurent, Godmer Alexandre
Team "Resist" UMR1184 "Immunology of Viral, Auto-Immune, Hematological and Bacterial diseases (IMVA-HB)", INSERM, Université Paris-Saclay, CEA, LabEx LERMIT, Faculty of Medicine, Le Kremlin-Bicêtre, France.
Bacteriology-Hygiene Unit, Bicêtre Hospital, Assistance Publique-Hôpitaux de Paris, Le Kremlin-Bicêtre, France.
J Mass Spectrom Adv Clin Lab. 2025 May 3;37:9-13. doi: 10.1016/j.jmsacl.2025.04.011. eCollection 2025 Aug.
The genus including clinically isolated species and , has a still underexplored role in clinical microbiology. Despite the clinical relevance of spp., current MALDI-TOF commercial systems fail to differentiate these species. Whole genome sequencing (WGS) remains the most effective method to distinguish species. However, this method is not adapted for routine lab workflow. Enhancing MALDI-TOF's accuracy could make it a rapid and effective approach for distinguishing species in routine laboratory diagnostics.
This study aims to improve the performance of MALDI-TOF for identifying spp. using WGS as the gold-standard reference method.
We applied Machine Learning (ML) algorithms to a collection of 235 clinicial spp. strains to develop an optimized identification model. Whole genome sequencing was used to characterize these strains and perform phylogenetic analysis, categorizing 209 strains as and 26 as .
The ML-based classifiers showed improved identification accuracy (44 of the 160 designed with accuracy at 1). Also, MS analysis identified 11 peaks able to discriminate between and .
Through development of a publicly-available online ML-based classifier, this study has improved the capacity of MALDI-TOF for distinguishing spp, providing a reliable, user-friendly solution suited to routine clinical diagnostics and supporting a better understanding of the roles of and in human pathology.
包括临床分离物种 和 在内的该属在临床微生物学中的作用仍未得到充分探索。尽管 属物种具有临床相关性，但当前的基质辅助激光解吸电离飞行时间质谱（MALDI-TOF）商业系统无法区分这些物种。全基因组测序（WGS）仍然是区分物种的最有效方法。然而，这种方法并不适用于常规实验室工作流程。提高MALDI-TOF的准确性可以使其成为在常规实验室诊断中区分 属物种的快速有效方法。
本研究旨在以WGS作为金标准参考方法，提高MALDI-TOF鉴定 属物种的性能。
我们将机器学习（ML）算法应用于235株临床 属物种菌株，以开发优化的鉴定模型。使用全基因组测序对这些菌株进行表征并进行系统发育分析，将209株菌株归类为 ，26株归类为 。
基于ML的分类器显示出提高的鉴定准确性（160个设计中的44个准确性为1）。此外，质谱分析确定了11个能够区分 和 的峰。
通过开发一个公开可用的基于ML的在线分类器，本研究提高了MALDI-TOF区分 属物种的能力，提供了一种适用于常规临床诊断的可靠、用户友好的解决方案，并有助于更好地理解 和 在人类病理学中的作用。
