Direct MALDI-TOF MS Identification of Bacterial Mixtures

Matrix-assisted
laser desorption/ionization time-of-flight mass
spectrometry (MALDI-TOF MS) is now widely used to characterize bacterial
samples for clinical diagnosis, food safety control, environmental
monitoring, and so on. However, existing standard approaches are only
applied to analyze single colonies purified by plate culture, which
limits the approaches to cultivable bacteria and makes the whole approaches
time-consuming. In this work, we propose a new framework to analyze
MALDI-TOF spectra of bacterial mixtures and to directly characterize
each component without purification procedures. The framework is a
combination of a synthetic mixture model based on a non-negative linear
combination of candidate reference spectra and a statistical assessment
by in silico generated spectra via a jackknife resampling. Ninety-seven
model bacterial mixture samples and 8 cocultured blind-coded bacterial
mixture samples, containing up to 6 strains in varied ratios in each
sample, together with a reference database containing the mass spectra
of 1081 strains, were used to validate the framework. High sensitivity
(>80%, with error rate <5%) was achieved for balanced binary
and
ternary mixtures. The sensitivity was >60% for balanced quaternary
and pentabasic mixtures, and 48%–71% for asymmetric situation,
with error rate <5%. The work can facilitate rapid and reliable
characterization of bacterial mixtures without purification procedures,
which is of practical value in clinical diagnosis, food safety control,
environmental monitoring, and so on. The framework can be further
applied to many other spectroscopy-based analytics to interpret spectra
from mixed samples.

基质辅助激光解吸电离飞行时间质谱（MALDI-TOF MS）目前已被广泛用于细菌样本的表征分析，应用场景涵盖临床诊断、食品安全管控、环境监测等诸多领域。然而，现有的标准分析方法仅适用于分析平板培养纯化得到的单菌落，这将方法的适用范围限制在可培养细菌上，同时导致整体分析流程耗时较长。本研究提出一种全新的分析框架，可直接对细菌混合样本的MALDI-TOF质谱数据进行解析，且无需纯化步骤即可直接表征其中的各组分。该框架结合了两大核心模块：一是基于候选参考质谱非负线性组合的合成混合样本模型，二是通过刀切法重采样生成计算机模拟质谱所开展的统计评估。研究共使用97个模拟细菌混合样本、8个共培养盲编码细菌混合样本（每个样本最多包含6株细菌，且各菌株占比各不相同），以及一个涵盖1081株细菌质谱数据的参考数据库，对该框架进行验证。对于均衡配比的二元及三元混合样本，该框架实现了超过80%的灵敏度，错误率低于5%；对于均衡配比的四元及五元混合样本，灵敏度超过60%；在非均衡配比场景下，灵敏度可达48%~71%，且错误率均低于5%。本研究提出的框架可实现无需纯化步骤的细菌混合样本快速可靠表征，在临床诊断、食品安全管控、环境监测等领域均具备实际应用价值。该框架还可进一步推广至诸多其他基于光谱的分析任务，用于解析混合样本的光谱数据。