Karisa M. Pierce

Data Analysis Methods

Data analysis methods are essential to transform chromatographic data into useful information. There is an ongoing evolution of data analysis methods as increasingly challenging applications need to be addressed, instrumentation advances occur, separation speeds increase, data dimensionality increases, and growing volumes of data are collected. As the traditional data analysis methods have been covered in prior chapters, this chapter focuses primarily on advanced data analysis methods known as chemometrics. Chemometrics are mathematical methods that glean useful information from chemical data, especially large volumes of complex data that are not amenable to traditional or manual analysis. Chemometrics fill several analytical roles. First, the ability to identify and quantify analytes at extremely low chromatographic resolution can be provided. Second, pattern recognition and classification are often implemented, where a chromatogram provides a chemical fingerprint. Third, groups of compounds may be quantified in concert using chemometric methods. Finally, another area of broad interest is retention prediction for separation optimization.

Review of chemometric analysis techniques for comprehensive two dimensional separations data

Comprehensive two-dimensional (2D) separations, such as comprehensive 2D gas chromatography (GC×GC), liquid chromatography (LC×LC), and related instrumental techniques, provide very large and complex data sets. It is often up to the software to assist the analyst in transforming these complex data sets into useful information, and that is precisely where the field of chemometric data analysis plays a pivotal role. Chemometric tools for comprehensive 2D separations are continually being developed and applied as researchers make significant advances in novel state-of-the-art algorithms and software, and as the commercial sector continues to provide user friendly chemometric software. In this review, we build upon previous reviews of this topic, by focusing primarily on advances that have been reported in the past five years. Most of the reports focus on instrumental platforms using GC×GC with either flame ionization detection (FID) or time-of-flight mass spectrometry (TOFMS) detection, or LC×LC with diode array absorbance detection (DAD). The review covers the following general topics: data preprocessing techniques, target analyte techniques, comprehensive nontarget analysis techniques, and software for chemometrics in multidimensional separations.

Comprehensive analysis of yeast metabolite GC×GC–TOFMS data: combining discovery-mode and deconvolution chemometric software

The first extensive study of yeast metabolite GC x GC-TOFMS data from cells grown under fermenting, R, and respiring, DR, conditions is reported. In this study, recently developed chemometric software for use with three-dimensional instrumentation data was implemented, using a statistically-based Fisher ratio method. The Fisher ratio method is fully automated and will rapidly reduce the data to pinpoint two-dimensional chromatographic peaks differentiating sample types while utilizing all the mass channels. The effect of lowering the Fisher ratio threshold on peak identification was studied. At the lowest threshold (just above the noise level), 73 metabolite peaks were identified, nearly three-fold greater than the number of previously reported metabolite peaks identified (26). In addition to the 73 identified metabolites, 81 unknown metabolites were also located. A Parallel Factor Analysis graphical user interface (PARAFAC GUI) was applied to selected mass channels to obtain a concentration ratio, for each metabolite under the two growth conditions. Of the 73 known metabolites identified by the Fisher ratio method, 54 were statistically changing to the 95% confidence limit between the DR and R conditions according to the rigorous Students t-test. PARAFAC determined the concentration ratio and provided a fully-deconvoluted (i.e. mathematically resolved) mass spectrum for each of the metabolites. The combination of the Fisher ratio method with the PARAFAC GUI provides high-throughput software for discovery-based metabolomics research, and is novel for GC x GC-TOFMS data due to the use of the entire data set in the analysis (640 MB x 70 runs, double precision floating point).

A principal component analysis based method to discover chemical differences in comprehensive two-dimensional gas chromatography with time-of-flight mass spectrometry (GC × GC-TOFMS) separations of metabolites in plant samples

Comprehensive two-dimensional gas chromatography combined with time-of-flight mass spectrometry (GCxGC-TOFMS) provides high resolution separations of complex samples with a mass spectrum at every point in the separation space. The large volumes of multidimensional data obtained by GCxGC-TOFMS analysis are analyzed using a principal component analysis (PCA) method described herein to quickly and objectively discover differences between complex samples. In this work, we submitted 54 chromatograms to PCA to automatically compare the metabolite profiles of three different species of plants, namely basil (Ocimum basilicum), peppermint (Mentha piperita), and sweet herb stevia (Stevia rebaudiana), where there were 18 chromatograms for each type of plant. The 54 scores of the m/z 73 data set clustered in three groups according to the three types of plants. Principal component 1 (PC 1) separated the stevia cluster from the basil and peppermint clusters, capturing 61.84% of the total variance. Principal component 2 (PC 2) separated the basil cluster from the peppermint cluster, capturing 16.78% of the total variance. The PCA method revealed that relative abundances of amino acids, carboxylic acids, and carbohydrates were responsible for differentiating the three plants. A brief list of the 16 most significant metabolites is reported. After PCA, the 54 scores of the m/z 217 data set clustered in three groups according to the three types of plants, as well, yielding highly loaded variables corresponding with chemical differences between plants that were complementary to the m/z 73 information. The PCA data mining method is applicable to all of the monitored selective mass channels, utilizing all of the collected data, to discover unknown differences in complex sample profiles.

Development toward rapid and efficient screening for high performance hydrolysate lots in a recombinant monoclonal antibody manufacturing process

Plant‐derived hydrolysates are widely used in mammalian cell culture media to increase yields of recombinant proteins and monoclonal antibodies (mAbs). However, these chemically varied and undefined raw materials can have negative impact on yield and/or product quality in large‐scale cell culture processes. Traditional methods that rely on fractionation of hydrolysates yielded little success in improving hydrolysate quality. We took a holistic approach to develop an efficient and reliable method to screen intact soy hydrolysate lots for commercial recombinant mAb manufacturing. Combined high‐resolution 1H nuclear magnetic resonance (NMR) spectroscopy and partial least squares (PLS) analysis led to a prediction model between product titer and NMR fingerprinting of soy hydrolysate with cross‐validated correlation coefficient R2 of 0.87 and root‐mean‐squared‐error of cross‐validation RMSECV% of 11.2%. This approach screens for high performance hydrolysate lots, therefore ensuring process consistency and product quality in the mAb manufacturing process. Furthermore, PLS analysis was successful in discerning multiple markers (DL‐lactate, soy saccharides, citrate and succinate) among hydrolysate components that positively and negatively correlate with titer. Interestingly, these markers correlate to the metabolic characteristics of some strains of taxonomically diverse lactic acid bacteria (LAB). Thus our findings indicate that LAB strains may exist during hydrolysate manufacturing steps and their biochemical activities may attribute to the titer enhancement effect of soy hydrolysates.

Predicting percent composition of blends of biodiesel and conventional diesel using gas chromatography-mass spectrometry, comprehensive two-dimensional gas chromatography-mass spectrometry, and partial least squares analysis.

The percent composition of blends of biodiesel and conventional diesel from a variety of retail sources were modeled and predicted using partial least squares (PLS) analysis applied to gas chromatography-total-ion-current mass spectrometry (GC-TIC), gas chromatography-mass spectrometry (GC-MS), comprehensive two-dimensional gas chromatography-total-ion-current mass spectrometry (GCxGC-TIC) and comprehensive two-dimensional gas chromatography-mass spectrometry (GCxGC-MS) separations of the blends. In all four cases, the PLS predictions for a test set of chromatograms were plotted versus the actual blend percent composition. The GC-TIC plot produced a best-fit line with slope=0.773 and y-intercept=2.89, and the average percent error of prediction was 12.0%. The GC-MS plot produced a best-fit line with slope=0.864 and y-intercept=1.72, and the average percent error of prediction was improved to 6.89%. The GCxGC-TIC plot produced a best-fit line with slope=0.983 and y-intercept=0.680, and the average percent error was slightly improved to 6.16%. The GCxGC-MS plot produced a best-fit line with slope=0.980 and y-intercept=0.620, and the average percent error was 6.12%. The GCxGC models performed best presumably due to the multidimensional advantage of higher dimensional instrumentation providing more chemical selectivity. All the PLS models used 3 latent variables. The chemical components that differentiate the blend percent compositions are reported.

Analysis of bacteria by pyrolysis gas chromatography–differential mobility spectrometry and isolation of chemical components with a dependence on growth temperature

Pyrolysis gas chromatography-differential mobility spectrometry (py-GC-DMS) analysis of E. coli, P. aeruginosa, S. warneri and M. luteus, grown at temperatures of 23, 30, and 37 degrees C, provided data sets of ion intensity, retention time, and compensation voltage for principal component analysis. Misaligned chromatographic axes were treated using piecewise alignment, the impact on the degree of class separation (DCS) of clusters was minor. The DCS, however, was improved between 21 to 527% by analysis of variance with Fisher ratios to remove chemical components independent of growth temperature. The temperature dependent components comprised 84% of all peaks in the py-GC-DMS analysis of E. coli and were attributed to the pyrolytic decomposition of proteins rather than lipids, as anticipated. Components were also isolated in other bacteria at differing amounts: 41% for M. luteus, 14% for P. aeruginosa, and 4% for S. warneri, and differing patterns suggested characteristic dependence on temperature of growth for these bacteria. These components are anticipated to have masses from 100 to 200 Da by inference from differential mobility spectra.

Predicting feedstock and percent composition for blends of biodiesel with conventional diesel using chemometrics and gas chromatography–mass spectrometry

The two main goals of the analytical method described herein were to (1) use principal component analysis (PCA), hierarchical clustering (HCA) and K-nearest neighbors (KNN) to determine the feedstock source of blends of biodiesel and conventional diesel (feedstocks were two sources of soy, two strains of jatropha, and a local feedstock) and (2) use a partial least squares (PLS) model built specifically for each feedstock to determine the percent composition of the blend. The chemometric models were built using training sets composed of total ion current chromatograms from gas chromatography-quadrupole mass spectrometry (GC-qMS) using a polar column. The models were used to semi-automatically determine feedstock and blend percent composition of independent test set samples. The PLS predictions for jatropha blends had RMSEC=0.6, RMSECV=1.2, and RMSEP=1.4. The PLS predictions for soy blends had RMSEC=0.5, RMSECV=0.8, and RMSEP=1.2. The average relative error in predicted test set sample compositions was 5% for jatropha blends and 4% for soy blends.

A Review of Chemometrics Applied to Comprehensive Two-dimensional Separations from 2008–2010

This review article covers developments in multidimensional separations combined with chemometrics that were published in 2008 through 2010, specifically for multidimensional gas chromatography, liquid chromatography, and electrophoresis. Although different instrumentation is used to generate multidimensional separations data, many similar data processing options and chemometrics can be applied in order to objectively distill the data into useful knowledge while reducing manual analysis and preserving data integrity. This review article describes the chemometrics employed in the referenced studies in terms of unsupervised, supervised, preprocessing, resolution, and image analysis algorithms. Other factors that affect converting data into useful knowledge are the structure of the data and the format of the data submitted to the analysis methods, so the studies are also described in terms of data dimensionality and data format (i.e., whether peak tables or raw data points were analyzed).

Chromatographic data analysis. Part 3.3.4: handling hyphenated data in chromatography

This tutorial describes how to analyze three dimensional (3D) pixel-level comprehensive two dimensional (2D) gas chromatography with time of flight mass spectrometry (GC × GC-TOFMS) data. The reader will learn how to baseline-correct and normalize the 3D pixel-level GC × GC-TOFMS chromatograms, how to make an n-way partial least squares (NPLS) model of thirteen 2D pixel-level total ion current (GC × GC-TIC) training set chromatograms, and how to use that NPLS model to predict the biodiesel blend percent composition of 9 independent test set chromatograms.