Objective Stool Classification

Leon Deutsch, Biotechnical Faculty, University of Ljubljana, Jamnikarjeva, Slovenia
Blaz Stres, Faculty of Civil and Geodetic Engineering, University of Ljubljana, Jamova, Slovenia & Department of Microbiology, University of Innsbruck, Austria

doi.org/10.3390/metabo11030172

Abstract

Past studies strongly connected stool consistency—as measured by Bristol Stool Scale (BSS)—with microbial gene richness and intestinal inflammation, colonic transit time and metabolome characteristics that are of clinical relevance in numerous gastro intestinal conditions. While retention time, defecation rate, BSS but not water activity have been shown to account for BSS-associated inflammatory effects, the potential correlation with the strength of a gel in the context of intestinal forces, abrasion, mucus imprinting, fecal pore clogging remains unexplored as a shaping factor for intestinal inflammation and has yet to be determined. Our study introduced a minimal pressure approach (MP) by probe indentation as measure of stool material crosslinking in fecal samples. Results reported here were obtained from 170 samples collected in two independent projects, including males and females, covering a wide span of moisture contents and BSS. MP values increased exponentially with increasing consistency (i.e., lower BSS) and enabled stratification of samples exhibiting mixed BSS classes. A trade-off between lowest MP and highest dry matter content delineated the span of intermediate healthy density of gel crosslinks. The crossectional transects identified fecal surface layers with exceptionally high MP and of <5 mm thickness followed by internal structures with an order of magnitude lower MP, characteristic of healthy stool consistency. The MP and BSS values reported in this study were coupled to reanalysis of the PlanHab data and fecal 1H-NMR metabolomes reported before. The exponential association between stool consistency and MP determined in this study was mirrored in the elevated intestinal and also systemic inflammation and other detrimental physiological deconditioning effects observed in the PlanHab participants reported before. The MP approach described in this study can be used to better understand fecal hardness and its relationships to human health as it provides a simple, fine scale and objective stool classification approach for the characterization of the exact sampling locations in future microbiome and metabolome studies.
Read more: The Importance of Objective Stool Classification in Fecal 1H-NMR Metabolomics

How was AutoML used?

Automated Machine Learning was used to identify most important metabolic features separating the groups of fecal matter plasticity. Samples were divided into three separate groups indicating low ((MP1 < 30), medium (30 < MP2 < 75), and high (MP3 > 75) MP, and 174 metabolic features were analyzed for possible differentiation. JADBIO version 1.1.182 was used for model generation. Data were split in a 70:30 ratio for model training (70%) and model validation (30%). An extensive tuning effort with six CPU cores was used to compute the most interpretable classification model, which was selected based on the area under the curve (AUC) metric among 168952 trained models. Different algorithms with different combinations of tuned parameters were used for feature selection (LASSO regression and test-budgeted statistically equivalent signature) and for prediction (ridge logistic regression, support vector machines, classification random forest and classification trees). Metabolic data were preprocessed with constant removal and standardized. LASSO feature selection (penalty = 1.0, lambda = 1.558e-01) was used for metabolite feature selection. The output of feature selection was used for obtaining the best interpretable model using the predictive ridge logistic regression algorithm (with hyper-parameter penalty equal to 1.0).
The machine learning process described in this study was adopted for several reasons: (i) automation in parameter and algorithm selection results in reduced bias and human interference; (ii) the approach includes several different ML algorithms (linear regression, SVM, decision tree, random forest and Gaussian kernel SVMs and automatically choose the most interpretable model based on AUC metric; (iii) the resulting models were trained with different configurations on different sub-samples of the original dataset (cross- val- idation); (iv) focus on relevant humanly interpretable models. Consequently, algorithm, hyperparameter and space selection (AHPS) as implemented in JADBIO was used for selecting the most suitable algorithm for preprocessing and transformation of a given dataset, its feature selection and modeling. The output of AHPS step was analyzed by configuration evaluation protocol (CEP) in order to find the optimal configuration reported in this study.