Phytothermotherapy (“grass baths”) is a traditional phytotherapy for rheumatism consisting of taking baths in hot fermenting grass. Scientific studies have demonstrated its efficiency in treating several rheumatic diseases. However the efficiency and repeatability of the therapy is dependent on the wild fermentations, determining sometimes the appearance of unpleasant conditions leading to the early abandonment of the therapy. The metabolism undergoing in the grass baths is unknown and there is not an established method to evaluate and predict grass baths quality.

The aim of this study is to establish a simple VOCs profiling method able to evaluate the grass baths, predicting their evolution, through the identification of marker volatiles related to the best conditions and/or the spoilage. After replicating in real scale the traditional grass baths, the volatile profiles were measured using passive diffusion samplers injected in a thermal desorption-comprehensive GC × GC-TOF-MS. The high dimensionality of the data coupled with the limited number of time points, required a rigorous method development for the analysis of the data, achieved through the development of a novel R package for variable selection in GC × GC data matrices. The further application of a fuzzy clustering approach demonstrated to be a useful tool dealing with short time series, allowing to discard un-trending volatiles and giving a clear snapshot of the main trends in the data. A broad coverage of the volatolome was provided, thus suitable to describe the main metabolic changes ongoing in the grass baths. Coupling this data with the temperature and pH, and comparing it to the data from similar processes, like silage and compost, we demonstrated that the established method can be helpful to evaluate short time series, allowing us to obtain a list of volatiles as candidate markers for the quality of the grass baths.

The established method gave a list of markers applicable to real scale grass baths to predict spoilage; furthermore it provides a list of volatiles where to search for candidate markers with reported health-related effects and can be used to generate hypothesis on the mechanisms of action of the treatment.

植物热疗法（“草浴”）是风湿病的传统植物疗法，包括在热发酵的草丛中洗澡。科学研究证明了其在治疗多种风湿性疾病中的功效。但是，该疗法的效率和可重复性取决于野生发酵，有时会确定令人不快的状况的出现，从而导致该疗法的早期放弃。在草浴中进行的新陈代谢是未知的，并且还没有建立评估和预测草浴质量的方法。

这项研究的目的是建立一种简单的VOC分析方法，通过鉴定与最佳条件和/或腐败相关的标志物挥发物，来评估草浴，预测草浴的演变。在实际复制传统草浴之后，使用注入到热解吸综合GC×GC-TOF-MS中的无源扩散采样器测量挥发性分布。数据的高维度性以及有限的时间点数量，要求开发严格的方法来分析数据，这是通过开发新颖的R包来进行GC×GC数据矩阵中的变量选择而实现的。事实证明，模糊聚类方法的进一步应用是处理短时间序列的有用工具，允许丢弃不趋势的挥发物，并提供数据主要趋势的清晰快照。提供了广泛的挥发性成分，因此适合描述草浴中正在进行的主要代谢变化。将这些数据与温度和pH值耦合，并将其与类似过程（如青贮饲料和堆肥）的数据进行比较，我们证明了所建立的方法可以帮助评估短时间序列，从而使我们能够获得一系列挥发物作为候选标记物对于草浴的质量。

建立的方法给出了适用于实际规模的草浴以预测变质的标志物列表；此外，它提供了挥发物清单，可在其中搜索具有已报告的与健康相关的作用的候选标记，并可用于生成有关治疗作用机制的假设。