当前位置: X-MOL 学术Int. J. Mass Spectrom. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Laboratory validation of rapid discrimination of single microbial cells via SPAMS with machine learning
International Journal of Mass Spectrometry ( IF 1.8 ) Pub Date : 2020-08-01 , DOI: 10.1016/j.ijms.2020.116340
Chaowu Liu , Guoqu Zeng , Mei Li , Zhen Zhou

Abstract Accurate and rapid identification of microbes is of critical importance in atmospheric effect and public health study. Conventional workflows are off-line and time-consuming and procedures are multi-faceted. Mass spectrometry (MS) can be an alternative but is limited by low efficiency as well as low reproducibility for spectrum profiles. We systematically investigated the feasibility of applying SPAMS for discrimination of microbes through machine learning method. The microbial cells were directly nebulized and inhaled without pretreatment, which produced rich ion contents. The MS spectra derived from 7658 single cells comprising 5 species were analyzed using both statistical analyses and machine leaning algorithm. ANOVA analysis indicated that a total of 129 distinct mass-to-charge (m/z) ions were obtained and detected with different abundance. Further hierarchical clustering analysis categorized the spectra into five groups precisely corresponding to the five species that revealed evident discrimination. In parallel, classification models were constructed via 3 different algorithms, and fine classification was maintained even the variables of ions were reduced. The model random forest containing 19 specific ions accurately classified independently prepared microbial cells and validated those not used for model construction with high accuracy. In conclusion, combining single cell mass spectrometry and machine learning method with experimental validation, that demonstrated to achieve rapid and reliable prediction of single microbial cells based on their ion profiles. This approach provided a laboratory standards evidence, which was critical importance towards the rapid and reliable discrimination of microbes from ambient air.

中文翻译:

使用机器学习通过 SPAMS 快速区分单个微生物细胞的实验室验证

摘要 准确、快速地识别微生物在大气效应和公共卫生研究中具有重要意义。传统的工作流程是离线且耗时的,而且程序是多方面的。质谱 (MS) 可以是一种替代方法,但受到效率低以及光谱图谱重现性低的限制。我们系统地研究了通过机器学习方法应用 SPAMS 来区分微生物的可行性。微生物细胞未经预处理直接雾化吸入,产生丰富的离子含量。使用统计分析和机器学习算法分析源自包含 5 个物种的 7658 个单细胞的 MS 光谱。方差分析表明,总共获得并检测到了 129 个不同的质荷比 (m/z) 离子,具有不同的丰度。进一步的层次聚类分析将光谱分为五组,精确对应于显示明显区分的五个物种。同时,通过3种不同的算法构建分类模型,即使减少了离子的变量,也能保持精细的分类。包含 19 个特定离子的模型随机森林对独立制备的微生物细胞进行了准确分类,并以高精度验证了未用于模型构建的微生物细胞。总之,将单细胞质谱和机器学习方法与实验验证相结合,这证明可以根据其离子谱对单个微生物细胞进行快速可靠的预测。这种方法提供了实验室标准证据,这对于从环境空气中快速可靠地区分微生物至关重要。
更新日期:2020-08-01
down
wechat
bug