Keshav Kumar

Discrete wavelet assisted correlation optimised warping of chromatograms: optimizing the computational time for correcting the drifts in peak positions

Correlation optimised warping (COW) has been the most favourite chromatographic peak alignment approach in recent years. After optimization of the two parameters, slack and segment length, COW work ...

Optimizing the process of reference selection for correlation optimised warping (COW) and interval correlation shifting (icoshift) analysis: automating the chromatographic alignment procedure

In order to ensure that the results of chromatographic data analysis workflow are well within the chemical and biological premise, the same chromatographic peak must be present at the same position for all the analysed samples. The correlation optimised warping (COW) and interval correlation shifting (icoshift) algorithm are the two most commonly used approaches that are used to correct the drifts in the peak position. Both the approaches work with different algorithms, COW works on expansion and compression approach whereas the icoshift works on an insertion and deletion approach. Both the approaches have their pros and cons. However, both the methods suffer with the problems associated with the selection of the proper reference sample. There are several ways of selecting a reference sample but none of them provides an unambiguous choice. Often, the selection of an inappropriate sample as the reference makes the whole alignment exercise a laborious task with very little success in correcting the drifts in peak position. An unambiguous reference selection procedure is the current prime requirement in the areas of proteomics, metabolomics, and the food industry that use chromatography-based analytical procedures. The present work addresses this issue by taking the advantage of the fact that chromatographic peaks can be approximated using the Gaussian function and proposes an approach that involves synthesis of a reference chromatogram using the Gaussian function for the subsequent COW and icoshift analysis. The proposed approach is successfully validated using the simulated as well as real life chromatograms and the results obtained are evaluated by means of several statistical parameters. The results clearly show that the proposed approach can reduce the ambiguity in the selection of a reference chromatogram and can speed up and automate the whole alignment procedure.

Eigenvalue-eigenvector decomposition (EED) analysis of dissimilarity and covariance matrix obtained from total synchronous fluorescence spectral (TSFS) data sets of herbal preparations: Optimizing the classification approach

The present work compares the dissimilarity and covariance based unsupervised chemometric classification approaches by taking the total synchronous fluorescence spectroscopy data sets acquired for the cumin and non-cumin based herbal preparations. The conventional decomposition method involves eigenvalue-eigenvector analysis of the covariance of the data set and finds the factors that can explain the overall major sources of variation present in the data set. The conventional approach does this irrespective of the fact that the samples belong to intrinsically different groups and hence leads to poor class separation. The present work shows that classification of such samples can be optimized by performing the eigenvalue-eigenvector decomposition on the pair-wise dissimilarity matrix.

Chemometric assisted correlation optimized warping of chromatograms: optimizing the computational time for correcting the drifts in chromatographic peak positions

Correlation optimized warping has been the most used technique to correct the drifts in peak positions. COW aligns the unaligned chromatogram to the reference chromatogram provided slack (t) and segment lengths (m) are optimized. However, several combinations of m and t need to be tested before finding the optimum combination of m and t. The optimization of these two parameters is laborious and computationally time consuming. The computational time significantly increases with the number of samples. As a result, often the correction of the retention time drifts becomes the bottleneck in the data analysis workflow. The present work shows that the application of constraint randomized non-negative factor analysis (CRNNFA), a chemometric technique, prior to COW can optimize the computational time by reducing the volume of datasets and subsequently allows a swift correction of drifts in peak positions. The utility of the proposed approach is successfully demonstrated by analyzing simulated as well as real life chromatograms. It is expected that this proposed approach will be useful and would be successfully integrated in the data analysis workflow related to research dealing with the finding of chemical, biochemical or clinical markers.

PG-metrics : a chemometric-based approach for classifying bacterial peptidoglycan data sets and uncovering their subjacent chemical variability

Bacteria cells are protected from osmotic and environmental stresses by an exoskeleton-like polymeric structure called peptidoglycan (PG) or murein sacculus. This structure is fundamental for bacteria’s viability and thus, the mechanisms underlying cell wall assembly and how it is modulated serve as targets for many of our most successful antibiotics. Therefore, it is now more important than ever to understand the genetics and structural chemistry of the bacterial cell walls in order to find new and effective methods of blocking it for the treatment of disease. In the last decades, liquid chromatography and mass spectrometry have been demonstrated to provide the required resolution and sensitivity to characterize the fine chemical structure of PG. However, the large volume of data sets that can be produced by these instruments today are difficult to handle without a proper data analysis workflow. Here, we present PG-metrics, a chemometric based pipeline that allows fast and easy classification of bacteria according to their muropeptide chromatographic profiles and identification of the subjacent PG chemical variability between e.g. bacterial species, growth conditions and, mutant libraries. The pipeline is successfully validated here using PG samples from different bacterial species and mutants in cell wall proteins. The obtained results clearly demonstrated that PG-metrics pipeline is a valuable bioanalytical tool that can lead us to cell wall classification and biomarker discovery.

Random Initialisation of the Spectral Variables: an Alternate Approach for Initiating Multivariate Curve Resolution Alternating Least Square (MCR-ALS) Analysis

Multivariate curve resolution alternating least square (MCR-ALS) analysis is the most commonly used curve resolution technique. The MCR-ALS model is fitted using the alternate least square (ALS) algorithm that needs initialisation of either contribution profiles or spectral profiles of each of the factor. The contribution profiles can be initialised using the evolve factor analysis; however, in principle, this approach requires that data must belong to the sequential process. The initialisation of the spectral profiles are usually carried out using the pure variable approach such as SIMPLISMA algorithm, this approach demands that each factor must have the pure variables in the data sets. Despite these limitations, the existing approaches have been quite a successful for initiating the MCR-ALS analysis. However, the present work proposes an alternate approach for the initialisation of the spectral variables by generating the random variables in the limits spanned by the maxima and minima of each spectral variable of the data set. The proposed approach does not require that there must be pure variables for each component of the multicomponent system or the concentration direction must follow the sequential process. The proposed approach is successfully validated using the excitation-emission matrix fluorescence data sets acquired for certain fluorophores with significant spectral overlap. The calculated contribution and spectral profiles of these fluorophores are found to correlate well with the experimental results. In summary, the present work proposes an alternate way to initiate the MCR-ALS analysis.

Chromatographic unsupervised classification of olive and non-olive oil samples with the aid of graph theory

Graph theory is a tool originating from discrete mathematics. A graph essentially consists of two fundamental units, nodes and edges. The nodes represent the samples and edges describe their connections. The edges are usually weighted with a dissimilarity value. Two nodes are similar if they have a smaller edge weight. In the present work, the analytical potential of graph theory and its ability to capture the heterogeneity present in the datasets are explored by analysing the high performance liquid chromatography (HPLC) datasets of 118 samples belonging to the classes of olive and non-olive oils. The graph theory based model clearly discriminated between the oil samples belonging to different classes. The obtained results show that graph theory can be used to achieve unsupervised classification of the samples. The present work suggests that graph theory should be considered as a useful analytical approach for analysing the data acquired for samples belonging to environmental, clinical, and pharmaceutical fields.

Chemometric assisted Fourier Transform Infrared (FTIR) Spectroscopic analysis of fruit wine samples: Optimizing the initialization and convergence criteria in the non-negative factor analysis algorithm for developing a robust classification model

The present work proposes certain optimization in the non-negative factor analysis (NNFA) algorithm to ensure an efficient analysis of the Fourier transformation infrared (FTIR) spectral data sets of the fruit wine samples. The first optimization deals with initialization of the variables in a controlled fashion that would ensure a reasonably good quality initial estimate to implement NNFA algorithm. It prevents NNFA algorithm from itinerating with random numbers that essentially have no chemical relevance. The second implemented optimization involves eliminating the alternate least square of convergence and allowing the algorithm to iterate until the iteration limit is reached. This criterion avoids the algorithm to have premature convergence and ensures that model provide the solutions which corresponds to the global minima. The application of NNFA with suggested optimizations are found to capture the subtle differences in the spectral profiles and classify the fruit wine samples that are essentially complex mixtures of several chemicals in unknown proportions. The proposed approach is also found to perform better than principal component analysis on practical grounds. In summary, the current work provides a simple, sensitive and cost-effective approach using optimized NNFA and FTIR spectroscopy for classifying the fruit wine samples.

Random initialisation of the excitation-emission matrix fluorescence spectral variables in constraint fashion for subsequent multivariate curve resolution alternating least square analysis on a peculiarly designed calibration set: Simultaneous sensing of nine polycyclic aromatic hydrocarbons in water samples

Polycyclic aromatic hydrocarbons (PAHs) are carcinogenic and mutagenic in nature therefore their sensing in water sample is an important analytical task. In the present work, a novel approach that is based on the random initialisation of the excitation-emission matrix fluorescence (EEMF) spectral variables in constraint fashion for subsequent multivariate curve resolution alternating least Square (MCR-ALS) analysis is introduced for simultaneously sensing the complex dilute aqueous mixture of PAHs. The usefulness of the proposed analytical approach is successfully demonstrated by applying it intentionally on a calibration set that is peculiar in many senses. The peculiarity mainly arises because the designed (i) the calibration set consist of nine PAHS having significant spectral overlap, (ii) the concentration of each PAH in different samples are kept constant and (iii) any two samples differ only in the presence and absence of the PAHs. The proposed approach is found to make precise and accurate estimation of each of the nine PAHs without involving any pre-separation. In summary, the proposed approach provides a simple and cost-effective procedure for simultaneous sensing of several PAHs in water samples. The proposed approach could be very useful in developing countries.

Processing Excitation-Emission Matrix Fluorescence and Total Synchronous Fluorescence Spectroscopy Data Sets with Constraint Randomised Non-negative Factor Analysis: a Novel Fluorescence Based Analytical Procedure to Analyse the Multifluorophoric Mixtures

The present work successfully shows the application of novel chemometric approach constraint randomised non-negative factor analysis (CRNNFA) for the analyses of the composite multidimensional fluorescence data sets. The CRNNFA involves the initialisation of the spectral variables in a constraint fashion thus ensures that algorithm does not wander with chemically and spectro-chemically irrelevant variables. The CRNNFA approach does not require that there must be pure variables for each fluorophores of the multifluorophoric mixture. One of the biggest advantages of CRNNFA is that it does not involve any convergence criteria thus circumventing the premature convergence of the algorithm. The CRNNFA achieves the termination only when the iteration limit is reached. The CRNNFA analysis s carried out under the non-negativity constraints therefore the mathematically retrieved profiles can easily be compared with those obtained experimentally. In the present work, both trilinear as well as non-trilinear multidimensional data sets are subjected to CRNNFA to validate its applicability. Excitation emission matrix fluorescence (EEMF) spectral profiles of Catechol, Hydroquinone, Indole and Tryptophan mixtures is used as the source of trilinear data sets. Total synchronous fluorescence spectroscopy (TSFS) spectral profiles of Benzo[a] Pyrene, Chrysene and Pyrene mixtures are used as the source of non-trilinear data sets. The CRNNFA approach is found to work equally well with trilinear as well with non-trilinear data sets. Thus, CRNFFA clearly does not have any prerequisite in the data structure. The obtained results clearly shows that CRNNFA algorithm in combination with EEMF and TSFS data sets are potential analytical tool for the analysis of complex-multifluorophoric mixtures.