Artificial Neural Network for Probabilistic Feature Recognition in Liquid Chromatography Coupled to High-Resolution Mass Spectrometry.

In this work, a novel probabilistic untargeted feature detection algorithm for liquid chromatography coupled to high-resolution mass spectrometry (LC-HRMS) using artificial neural network (ANN) is presented. The feature detection process is approached as a pattern recognition problem, and thus, ANN was utilized as an efficient feature recognition tool. Unlike most existing feature detection algorithms, with this approach, any suspected chromatographic profile (i.e., shape of a peak) can easily be incorporated by training the network, avoiding the need to perform computationally expensive regression methods with specific mathematical models. In addition, with this method, we have shown that the high-resolution raw data can be fully utilized without applying any arbitrary thresholds or data reduction, therefore improving the sensitivity of the method for compound identification purposes. Furthermore, opposed to existing deterministic (binary) approaches, this method rather estimates the probability of a feature being present/absent at a given point of interest, thus giving chance for all data points to be propagated down the data analysis pipeline, weighed with their probability. The algorithm was tested with data sets generated from spiked samples in forensic and food safety context and has shown promising results by detecting features for all compounds in a computationally reasonable time.

Bayesian peak tracking: A novel probabilistic approach to match GCxGC chromatograms.

A novel peak tracking method based on Bayesian statistics is proposed. The method consists of assigning (i.e. tracking) peaks from two GCxGC-FID data sets of the same sample taken in different conditions. Opposed to traditional (i.e. deterministic) peak tracking algorithms, in which the assignment problem is solved with a unique solution, the proposed algorithm is probabilistic. In other words, we quantify the uncertainty of matching two peaks without excluding other possible candidates, ranking the possible peak assignments regarding their posterior probability. This represents a significant advantage over existing deterministic methods. Two algorithms are presented: the blind peak tracking algorithm (BPTA) and peak table matching algorithm (PTMA). PTMA method was able to assign correctly 78% of a selection of peaks in a GCxGC-FID chromatogram of a diesel sample and proved to be extremely fast.

Exhaled breath profiling in diagnosing wheezy preschool children.

Although wheeze is common in preschool children, the underlying pathophysiology has not yet been disentangled. Volatile organic compounds (VOCs) in exhaled breath may serve as noninvasive markers of early wheeze. We aimed to assess the feasibility of VOC collection in preschool children, and to study whether a VOC profile can differentiate between children with and without recurrent wheeze. We included children (mean (range) age 3.3 (1.9-4.5) yrs) with (n=202) and without (n=50) recurrent wheeze. Exhaled VOCs were analysed by gas chromatography-time-of-flight mass spectrometry. VOC profiles were generated by ANOVA simultaneous component analysis (ASCA) and sparse logistic regression (SLR). Exhaled breath collection was possible in 98% of the children. In total, 913 different VOCs were detected. The signal-to-noise ratio improved after correction for age, sex and season using ASCA pre-processing. An SLR model with 28 VOCs correctly classified 83% of the children (84% sensitivity, 80% specificity). After six-fold cross-validation, 73% were correctly classified (79% sensitivity, 50% specificity). Assessment of VOCs in exhaled breath is feasible in young children. VOC profiles are able to distinguish children with and without recurrent wheeze with a reasonable accuracy. This proof of principle paves the way for additional research on VOCs in preschool wheezing.

Alignment and clustering strategies for GC×GC-MS features using a cylindrical mapping.

Comprehensive two-dimensional gas chromatography coupled to mass spectrometry is a powerful tool to analyze complex samples. For application of the technique in studies like biomarker discovery in which large sets of complex samples have to be analyzed, extensive preprocessing is needed to align the data obtained in several injections (analyses). We developed new alignment and clustering algorithms for this type of data. New in the current procedures is the consistent way in which the phenomenon referred to as wrap-around is treated. The data analysis problems associated with this phenomenon are solved by treating the 2D display as the surface of a three-dimensional cylinder. Based on this transformation we developed a new similarity metric for features as a function of both the cylindrical distance (reflecting similarity in chromatographic behavior) and of the mass spectral correlation (reflecting similarity in chemical structure). The concepts are used in warping and clustering, and include a protection against greedy warping. The methods were applied - for the purpose of an example - to the analysis of 11 replicates of a human urine sample concentrated by solid phase extraction. It is shown that the alignment is well protected against greedy warping which is important with respect to analytical qualities as robustness and repeatability. It is also demonstrated that chemically similar features are clustered together. The paper is organized as follows. First a brief introduction is provided addressing the background of the GC×GC-MS data structure followed by a theoretical section with a conceptual description of the procedures and details of the algorithms. Finally an example is given in the experimental section, illustrating the application of the procedures.

Assessing the performance of statistical validation tools for megavariate metabolomics data.

Statistical model validation tools such as cross-validation, jack-knifing model parameters and permutation tests are meant to obtain an objective assessment of the performance and stability of a statistical model. However, little is known about the performance of these tools for megavariate data sets, having, for instance, a number of variables larger than 10 times the number of subjects. The performance is assessed for megavariate metabolomics data, but the conclusions also carry over to proteomics, transcriptomics and many other research areas. Partial least squares discriminant analyses models were built for several LC-MS lipidomic training data sets of various numbers of lean and obese subjects. The training data sets were compared on their modelling performance and their predictability using a 10-fold cross-validation, a permutation test, and test data sets. A wide range of cross-validation error rates was found (from 7.5% to 16.3% for the largest trainings set and from 0% to 60% for the smallest training set) and the error rate increased when the number of subjects decreased. The test error rates varied from 5% to 50%. The smaller the number of subjects compared to the number of variables, the less the outcome of validation tools such as cross-validation, jack-knifing model parameters and permutation tests can be trusted. The result depends crucially on the specific sample of subjects that is used for modelling. The validation tools cannot be used as warning mechanism for problems due to sample size or to representativity of the sampling.