G ‐scores: A method for identifying disease‐causing pathogens with application to lower respiratory tract infections

Abstract

Lower respiratory tract infections (LRTIs) are well known for the lack of a good diagnostic method. The main difficulty lies in the fact that there are a variety of pathogens causing LRTIs, and their management and treatment are quite different. The development of quantitative real‐time loop‐mediated isothermal amplification (qrt‐LAMP) made it possible to rapidly amplify and quantify multiple pathogens simultaneously. The question that remains to be answered is how accurate and reliable is this method? More importantly, how are qrt‐LAMP measurements utilized to inform/suggest medical decisions? When does a pathogen start to grow out of control and cause infection? Answers to these questions are crucial to advise treatment guidance for LRTIs and also helpful to design phase I/II trials or adaptive treatment strategies. In this article, our main contributions include the following two aspects. First, we utilize zero‐inflated mixture models to provide statistical evidence for the validity of qrt‐LAMP being used in detecting pathogens for LRTIs without the presence of a gold standard test. Our results on qrt‐LAMP suggest that it provides reliable measurements on pathogens of interest. Second, we propose a novel statistical approach to identify disease‐causing pathogens, that is, distinguish the pathogens that colonize without causing problems from those that rapidly grow and cause infection. We achieve this by combining information from absolute quantities of pathogens and their symbiosis information to form G ‐scores. Change‐point detection methods are utilized on these G ‐scores to detect the three phases of bacterial growth—lag phase, log phase, and stationary phase. Copyright © 2014 John Wiley & Sons, Ltd.