Grouping of complex substances using analytical chemistry data: A framework for quantitative evaluation and visualization

Abstract

A detailed characterization of the chemical composition of complex substances, such as products of petroleum refining and environmental mixtures, is greatly needed in exposure assessment and manufacturing. The inherent complexity and variability in the composition of complex substances obfuscate the choices for their detailed analytical characterization. Yet, in lieu of exact chemical composition of complex substances, evaluation of the degree of similarity is a sensible path toward decision-making in environmental health regulations. Grouping of similar complex substances is a challenge that can be addressed via advanced analytical methods and streamlined data analysis and visualization techniques. Here, we propose a framework with unsupervised and supervised analyses to optimally group complex substances based on their analytical features. We test two data sets of complex oil-derived substances. The first data set is from gas chromatography-mass spectrometry (GC-MS) analysis of 20 Standard Reference Materials representing crude oils and oil refining products. The second data set consists of 15 samples of various gas oils analyzed using three analytical techniques: GC-MS, GC×GC-flame ionization detection (FID), and ion mobility spectrometry-mass spectrometry (IM-MS). We use hierarchical clustering using Pearson correlation as a similarity metric for the unsupervised analysis and build classification models using the Random Forest algorithm for the supervised analysis. We present a quantitative comparative assessment of clustering results via Fowlkes–Mallows index, and classification results via model accuracies in predicting the group of an unknown complex substance. We demonstrate the effect of (i) different grouping methodologies, (ii) data set size, and (iii) dimensionality reduction on the grouping quality, and (iv) different analytical techniques on the characterization of the complex substances. While the complexity and variability in chemical composition are an inherent feature of complex substances, we demonstrate how the choices of the data analysis and visualization methods can impact the communication of their characteristics to delineate sufficient similarity.

Grouping of UVCB substances with new approach methodologies (NAMs) data

“…To evaluate the outcome of such grouping, we included a quantitative metric into the unsupervised analysis workflow to assess the correspondence of the outcome to the original categories of each chemical. The details of the unsupervised analysis workflow are described elsewhere ( Onel et al, 2019 ). Briefly, clustering was performed using the hclust function in R, using average linkage clustering applied to a Euclidean distance metric on centered, scaled data (essentially Pearson correlation), which we have previously found to be reasonably robust ( Onel et al, 2019 ).…”

“…The details of the unsupervised analysis workflow are described elsewhere ( Onel et al, 2019 ). Briefly, clustering was performed using the hclust function in R, using average linkage clustering applied to a Euclidean distance metric on centered, scaled data (essentially Pearson correlation), which we have previously found to be reasonably robust ( Onel et al, 2019 ). The Fowlkes-Mallows (FM) index ( Fowlkes and Mallows, 1983 ), a measure of similarity of two clusters, was calculated to enable quantitative comparative assessment between groupings achieved using each dataset to the known chemical categories.…”

“…Complementary approaches to grouping of petroleum-based UVCBs have been proposed based on in vitro testing ( Grimm et al, 2016 ; Kamelia et al, 2019 ), structure-activity analysis ( Kutsarova et al, 2019 ; Dimitrov et al, 2015 ), or novel analytical chemistry methods ( Grimm et al, 2017 ; Onel et al, 2019 ). Similar approaches have been applied to other categories of UVCBs ( Catlin et al, 2018 ).…”

A data‐driven optimization algorithm for differential algebraic equations with numerical infeasibilities

“…In machine learning, SVMs are extensively used for classification and regression-type of analyses, spanning over several different application areas including but not limited to fault detection and diagnosis, [30][31][32] improvement of process operations, 33 and predictive modeling of complex substances. 34,35 In this work, we use an SVM model to mimic the implicit constraint that defines the feasibility of the solution of a DAE system. Specifically, we build an SVM-based classification model in the offline phase by using a dataset of simulated samples with their binary outcome (feasible/infeasible).…”

Classification of estrogenic compounds by coupling high content analysis and machine learning algorithms

“…In this work, the training set is a balanced subset of the active compounds that contain both agonist and antagonist chemicals and their corresponding 4 technical replicates, whereas the testing set is the remaining unseen active compounds with their corresponding 4 technical replicates, not used in the training phase. Supervised learning algorithms are widely studied in many fields of engineering and sciences primarily in classification and regression-type problems for predicting either a categorical output or a continuous output, respectively [24,[26][27][28][29][30][31][32]. Classification is the problem of finding the categorical output of a new observation and distinguishing between different classes of information via statistical recognition of patterns in a training data set.…”