Multiplex protein analysis and ensemble machine learning methods of fine needle aspirates from prostate cancer patients reveal potential diagnostic signatures associated with tumour grade.

BACKGROUND: Improved molecular diagnosis is needed in prostate cancer (PC). Fine needle aspiration (FNA) is a minimally invasive biopsy technique, less traumatic compared to core needle biopsy, and could be useful for diagnosis of PC. Molecular biomarkers (BMs) in FNA-samples can be assessed for prediction, eg of immunotherapy efficacy before treatment as well as at treatment decision time points during disease progression. METHODS: In the present pilot study, the expression levels of 151 BM proteins were analysed by proximity extension assay in FNA-samples from 16 patients, including benign prostate lesions (n = 3) and cancers (n = 13). An ensemble data analysis strategy was applied using several machine learning models. RESULTS: Twelve potentially predictive BM proteins correlating with International Society of Urological Pathology grade groups were identified, among them vimentin, tissue factor pathway inhibitor 2, and integrin beta-5. The validity of the results was supported by network analysis that showed functional associations between most of the identified putative BMs. We also showed that multiple immune checkpoint targets can be assessed (eg PD-L1, CD137, and Galectin-9), which may support the selection of immunotherapy in advanced PC. Results are promising but need further validation in a larger cohort. CONCLUSIONS: Our pilot study represents a "proof of concept" and shows that multiplex profiling of potential diagnostic and predictive BM proteins is feasible on tumour material obtained by FNA sampling of prostate cancer. Moreover, our results demonstrate that an ensemble data analysis strategy may facilitate the identification of BM signatures in pilot studies when the patient cohort is limited.

Development of a high performance anion exchange chromatography analysis for mapping of oligosaccharides.

In the present study a HPAEC-PAD method is described that was developed for monitoring the consistency of N-glycosylation during the production and purification of recombinant proteins and monoclonal antibodies. The method successfully separated 18 neutral and sialylated oligosaccharides. Results obtained were compared with MALDI-TOF MS and it was shown that both methods gave similar results. In addition, a method validation was performed showing that the HPAEC-PAD analysis was well suited for the mapping and characterization of oligosaccharides. The method was found to be robust and additionally the precision was significantly better compared to the MALDI-TOF MS method.

Discrimination among IgG1-kappa monoclonal antibodies produced by two cell lines using charge state distributions in nanoESI-TOF mass spectra.

Charge state distributions (CSDs) of proteins in nanoESI mass spectra are affected by the instrumental settings and experimental conditions, in addition to the conformations of the proteins in the analyzed solutions. In the presented study, instrumental and experimental parameters--the desolvation gas flow rate, temperature, pH, buffer (ammonium acetate), and organic modifier (methanol) concentrations--were optimized according to a reduced central composite face experimental design to maximize the separation of CSDs of monoclonal IgG1-kappa antibodies produced by two production systems (CHO and GS-NS0 cell lines). Principal component analysis and Fisher linear discriminant analysis were then used to reduce the dimensions of the acquired dataset and quantify the separation of the protein classes, respectively. The results show that the IgG1-kappa molecules produced by the two production systems can be clearly distinguished using the described approach, which could be readily applied to other proteins and production systems.

Conformational studies of a monoclonal antibody, IgG1, by chemical oxidation: structural analysis by ultrahigh-pressure LC-electrospray ionization time-of-flight MS and multivariate data analysis.

We describe the development of a method in which protein oxidation by H2O2 followed by ultrahigh-pressure liquid chromatography (UHPLC) coupled with electrospray ionization time-of-flight mass spectrometry (ESI-ToFMS) and multivariate analysis are used to detect alterations in conformational states of proteins. In the study reported here, an IgG1 monoclonal antibody in native and denatured conformational states was oxidized by treatment with hydrogen peroxide. Peptide fragments generated by tryptic digestion were then analyzed by UHPLC-ESI-ToFMS. After reducing noise and extracting peaks from the LC-MS data using MzExplorer, software developed in-house and based on Matlab, we were able to distinguish peptides arising from the native and denatured states of the oxidized protein by principal component analysis. Peptides containing residues, which are inclined to undergo oxidation, such as methionine, are founded to be particularly important in this approach. We believe that the methodology could facilitate attempts to characterize the conformational states of recombinant monoclonal antibodies and other proteins.

Second-order peak detection for multicomponent high-resolution LC/MS data.

The first step when analyzing multicomponent LC/MS data from complex samples such as biofluid metabolic profiles is to separate the data into information and noise via, for example, peak detection. Due to the complex nature of this type of data, with problems such as alternating backgrounds and differing peak shapes, this can be a very complex task. This paper presents and evaluates a two-dimensional peak detection algorithm based on raw vector-represented LC/MS data. The algorithm exploits the fact that in high-resolution centroid data chromatographic peaks emerge flanked with data voids in the corresponding mass axis. According to the proposed method, only 4 per thousand of the total amount of data from a urine sample is defined as chromatographic peaks; however, 94% of the raw data variance is captured within these peaks. Compared to bucketed data, results show that essentially the same features that an experienced analyst would define as peaks can automatically be extracted with a minimum of noise and background. The method is simple and requires a priori knowledge of only the minimum chromatographic peak width-a system-dependent parameter that is easily assessed. Additional meta parameters are estimated from the data themselves. The result is well-defined chromatographic peaks that are consistently arranged in a matrix at their corresponding m/z values. In the context of automated analysis, the method thus provides an alternative to the traditional approach of bucketing the data followed by denoising and/or one-dimensional peak detection. The software implementation of the proposed algorithm is available at http://www.anchem.su.se/peakd as compiled code for Matlab.

Comparison of recent methods for inference of variable influence in neural networks.

Neural networks (NNs) belong to 'black box' models and therefore 'suffer' from interpretation difficulties. Four recent methods inferring variable influence in NNs are compared in this paper. The methods assist the interpretation task during different phases of the modeling procedure. They belong to information theory (ITSS), the Bayesian framework (ARD), the analysis of the network's weights (GIM), and the sequential omission of the variables (SZW). The comparison is based upon artificial and real data sets of differing size, complexity and noise level. The influence of the neural network's size has also been considered. The results provide useful information about the agreement between the methods under different conditions. Generally, SZW and GIM differ from ARD regarding the variable influence, although applied to NNs with similar modeling accuracy, even when larger data sets sizes are used. ITSS produces similar results to SZW and GIM, although suffering more from the 'curse of dimensionality'.

Metabolic fingerprinting of rat urine by LC/MS Part 2. Data pretreatment methods for handling of complex data.

Metabolic fingerprinting of biofluids like urine is a useful technique for detecting differences between individuals. With this approach, it might be possible to classify samples according to their biological relevance. In Part 1 of this work a method for the comprehensive screening of metabolites was described, using two different liquid chromatography (LC) column set-ups and detection by electrospray ionization mass spectrometry (ESI-MS). Data pretreatment of the resulting data described in is needed to reduce the complexity of the data and to obtain useful metabolic fingerprints. Three different approaches, i.e., reduced dimensionality (RD), MarkerLynx, and MS Resolver, were compared for the extraction of information. The pretreated data were then subjected to multivariate data analysis by partial least squares discriminant analysis (PLS-DA) for classification. By combining two different chromatographic procedures and data analysis, the detection of metabolites was enhanced as well as the finding of metabolic fingerprints that govern classification. Additional potential biomarkers or xenobiotic metabolites were detected in the fraction containing highly polar compounds that are normally discarded when using reversed-phase liquid chromatography.

Cross-linking of gelatin capsules with formaldehyde and other aldehydes: an FTIR spectroscopy study.

Attenuated total reflection Fourier transform infrared spectroscopy (FTIR) has been used to study cross-linking in hard gelatin capsules induced by exposure to formaldehyde, acetaldehyde, and propionaldehyde. These aldehydes are known to cause cross-linking between the amino acid chains of gelatin. Using FTIR spectroscopy, it is possible to analyze the cross-linking mechanisms by studying changes in the vibrational bands of the gelatin spectrum. The FTIR spectrum changes over time when the capsules are left in an aldehyde-rich environment. Analysis of the spectra shows that the early observed spectral changes conform to reaction intermediates proposed in previous work based on nuclear magnetic resonance experiments, specifically, the formation of amine methyl alcohol of arginine and lysine residues. Further spectral changes appear to be mostly from unreacted aldehydes absorbed to the gelatin, although a minor shift of the amide II peak is attributed to cross-link formation.

Metabolic fingerprinting of rat urine by LC/MS Part 1. Analysis by hydrophilic interaction liquid chromatography-electrospray ionization mass spectrometry.

Complex biological samples, such as urine, contain a very large number of endogenous metabolites reflecting the metabolic state of an organism. Metabolite patterns can provide a comprehensive signature of the physiological state of an organism as well as insights into specific biochemical processes. Although the metabolites excreted in urine are commonly highly polar, the samples are generally analyzed using reversed-phase liquid chromatography mass spectrometry (RP-LC/MS). In Part 1 of this work, a method for detecting highly polar metabolites by hydrophilic interaction liquid chromatography-electrospray ionization mass spectrometry (HILIC/ESI-MS) is described as a complement to RP-LC/ESI-MS. In addition, in an accompanying paper (Part 2), different multivariate approaches to extracting information from the resulting complex data are described to enable metabolic fingerprints to be obtained. The coverage of the method for the screening of as many metabolites as possible is highly improved by analyzing the urine samples using both a C(18) column and a ZIC-HILIC column. The latter was found to be a good alternative when analyzing highly polar compounds, e.g., hydroxyproline and creatinine, to columns typically used for reversed-phase liquid chromatography.

A comparison of methods for alignment of NMR peaks in the context of cluster analysis.

This paper compares the performance of two recently developed algorithms and methods for peak alignment of first-order NMR data of complex biological samples. The NMR spectra of such samples exhibit variations in peak position and peak shape due to variations in the sample matrix and to instrumental instabilities. The first method comprises an alignment of spectral segments with linear interpolation and shift correction to accommodate correspondence between a target and a test spectrum by a beam search or genetic algorithm. The second method is based on peak picking and needle vector representation of the NMR data with subsequent breadth-first search to establish shift corrections between the target and the test spectrum. The two proposed peak alignment methods and their respective merits are discussed for a real metabonomics application. Both alignment methods have been shown to enhance the interpretability of the resulting multivariate models, thereby increasing the prospect of detecting and following the onset of subtle biological changes reflected in the NMR data.

Determination of solvation free energies by adaptive expanded ensemble molecular dynamics.

A new method of calculating absolute free energies is presented. It was developed as an extension to the expanded ensemble molecular dynamics scheme and uses probability density estimation to continuously optimize the expanded ensemble parameters. The new method is much faster as it removes the time-consuming and expertise-requiring step of determining balancing factors. Its efficiency and accuracy are demonstrated for the dissolution of three qualitatively very different chemical species in water: methane, ionic salts, and benzylamine. A recently suggested optimization scheme by Wang and Landau [Phys. Rev. Lett. 86, 2050 (2001)] was also implemented and found to be computationally less efficient than the proposed adaptive expanded ensemble method.

Microemulsion electrokinetic chromatography of drugs varying in charge and hydrophobicity Part II: Strategies for optimization of separation.

The separation of anionic, cationic, and neutral drugs in microemulsion electrokinetic chromatography (MEEKC) was studied. The concentration of sodium dodecyl sulfate (SDS; surfactant) and 2-propanol (organic solvent) was varied in a three-level full factorial design. 29 different model substances were chosen with different hydrophobicities and charges (neutral, positive, and negative). The models were calculated by means of multiple linear regression (MLR). The compounds were divided into five different subgroups, and different strategies for optimization of the separation within each group were investigated. The optimization was done by maximizing the selectivity using response surface plots in MODDE, by calculation of different chromatographic functions, and by using the software DryLab. For all the different groups, MODDE, almost all chromatographic functions and DryLab gave approximately the same settings of the factors for optimum separation. Attempts were made to fit descriptors of the compounds to the retention data from the three-level full factorial design by means of partial least squares projection to latent structures (PLS). Between 86 and 89% of all predictions of migration times were acceptable (80-120% of the observed value).

Data preprocessing by wavelets and genetic algorithms for enhanced multivariate analysis of LC peptide mapping.

Peptide mapping by means of liquid chromatography is a powerful technique used for the characterisation and analysis of the primary structure of proteins. Subtle changes in the covalent structure of the protein can be detected by means of the chromatographic profile (fingerprint). Chromatographic methods, however, display variations in the chromatographic profile even at identical instrumental settings and sample conditions. These variations may be due to changes of the chromatographic conditions, e.g. slight shifts in column temperature, and degradation or alterations of the stationary phase or small changes in the trifluoroacetic acid (TFA) concentration. Such variations may result in varying retention times and peak shapes of the analytes and differences in the chromatographic baseline, thereby having a detrimental impact on the results obtained on multivariate analysis of peptide maps. In order to reduce the non-sample-related variations and to be able to more fully extract the information in peptide mapping, approaches for achieving this objective are outlined in the present study. These methods are denoising and data compression of the chromatograms by wavelets, baseline corrections by linear interpolation, and peak shift alignments towards a target chromatogram by means of a genetic algorithm. Visual inspections of preprocessed chromatograms and principal component analysis (PCA) score plots demonstrate the efficiency of the methodology used. Furthermore, deliberately added changes, e.g. insertions of small Gaussian peaks (outliers), are more easily detected by the proposed methods than from the original chromatograms by multivariate analysis.

Multivariate approaches for efficient detection of potential metabolites from liquid chromatography/mass spectrometry data.

This work describes a novel method for rapid screening of unknown metabolites in urine samples that narrows down the list of potential metabolites. Prior to analysis by liquid chromatography/electrospray ionization mass spectrometry (LC/ESI-MS), urine samples were prepared using solid-phase extraction (SPE). Automatic curve resolution was used for deconvolution of the LC/MS data, followed by peak alignment. Preprocessed data were then used for metabolite pattern recognition using principal component analysis (PCA), parallel factor analysis (PARAFAC), and multilinear partial least squares (N-PLS). This approach enabled the rapid detection of metabolites of citalopram in urine by maximizing the information extracted. The metabolites thus identified were compared with earlier studies on the metabolism of citalopram. In addition, new, unreported metabolites were found and characterized by LC/MS/MS and accurate mass measurements. A combination of data from positive and negative ionization enhanced the identification of metabolites.

Peak purity determination with principal component analysis of high-performance liquid chromatography-diode array detection data.

A method is proposed for the determination of chromatographic peak purity by means of principal component analysis (PCA) of high-performance liquid chromatography with diode array detection (HPLC-DAD) data. The method is exemplified with analysis of binary mixtures of lidocaine and prilocaine with different levels of separation. Lidocaine and prilocaine have very similar spectra and the chromatograms used had substantial peak overlap. The samples analysed contained a constant amount of lidocaine and a minor amount of prilocaine (0.02-2 conc.%) and hence the focus was on determining the purity of the lidocaine peak in the presence of much smaller levels of prilocaine. The peak purity determination was made by examination of relative observation residuals, scores and loadings from the PCA decomposition of DAD data over a chromatographic peak. As a reference method, the functions for peak purity analysis in the chromatographic data system used (Chromeleon) were applied. The PCA method showed good results at the same level as the detection limit of baseline-separated prilocaine, outperforming the methods in Chromeleon by a factor of ten. There is a discussion of the interpretation of the result, with some comparisons with evolving factor analysis (EFA). The main advantage of the PCA method for determination of peak purity over methods like EFA lies in its simplicity, short time of calculation and ease of use.

Microemulsion electrokinetic chromatography of drugs varying in charge and hydrophobicity: I. Impact of parameters on separation performance evaluated by multiple linear regression models.

The separation of anionic, cationic and neutral drugs in microemulsion electrokinetic chromatography (MEEKC) was studied with a statistical experimental design. The concentration of sodium dodecyl sulfate (SDS, surfactant), 1-butanol (co-surfactant) and borate buffer and the factors Brij 35 (surfactant), 2-propanol (organic solvent) and cassette temperature were varied simultaneously, while the parameters pH (9.2), the concentration of octane (oil, 0.8% w/w), the voltage (10 kV) and the dimension of the fused-silica capillary, were kept constant. Eight different model substances were chosen with different hydrophobicities. Two of the analytes were positively charged, two were negatively charged, and the remaining four were neutral or close to neutral at the pH explored. The importance of each parameter on the separation window, the plate height and the retention factor for each of the analytes was studied by means of multiple linear regression (MLR) models. A new response was evaluated for anions, the quotient between the effective mobility in the microemulsion and the effective mobility in the corresponding buffer. Factors affecting selectivity changes were also explored, and it was found that SDS and 2-propanol had the largest effect on selectivity.

Simultaneous determination of albumin and immunoglobulin G with fluorescence spectroscopy and multivariate calibration.

A method is proposed for the simultaneous determination of albumin and immunoglobulin G (IgG1) with fluorescence spectroscopy and multivariate calibration with partial least squares regression (PLS). The influence of some instrumental parameters were investigated with two experimental designs comprising 19 and 11 experiments, respectively. The investigated parameters were excitation and emission slit, detection voltage and scan rate. When a suitable instrumental setting had been found, a minor calibration and test set were analysed and evaluated. Thereafter, a larger calibration of albumin and IgG1 was made out of 26 samples (0-42mugml(-1) albumin and 0-12.7mugml(-1) IgG1). This calibration was validated with a test set consisting of 14 samples in the same concentration range. The precision of the method was estimated by analysing two test set samples for six times each. The scan modes tested were emission scan and synchronous scan Delta60nm. The results showed that the method could be used for determination of albumin and IgG1 (albumin, root mean square error of prediction (RMSEP) <2, relative standard error of prediction (RSEP) <6% and IgG1, RMSEP <1, RSEP <8%) in spite of the overlapping fluorescence of the two compounds. The estimated precision was relative standard deviation (R.S.D.) <1.7%. The method was finally applied for the analysis of some sample fractions from an albumin standard used in affinity chromatography.

Screening of biomarkers in rat urine using LC/electrospray ionization-MS and two-way data analysis.

Biofluids, like urine, form very complex matrixes containing a large number of potential biomarkers, that is, changes of endogenous metabolites in response to xenobiotic exposure. This paper describes a fast and sensitive method of screening biomarkers in rat urine. Biomarkers for phospholipidosis, induced by an antidepressant drug, were studied. Urine samples from rats exposed to citalopram were analyzed using solid-phase extraction (SPE) and liquid chromatography mass spectrometry (LC/MS) analysis detecting negative ions. A fast iterative method, called Gentle, was used for the automatic curve resolution, and metabolic fingerprints were obtained. After peak alignment principal component analysis (PCA) was performed for pattern recognition, PCA loadings were studied as a means of discovering potential biomarkers. In this study a number of potential biomarkers of phospholipidosis in rats are discussed. They are reported by their retention time and base peak, as their identification is not within the scope of the study. In addition to the fact that it was possible to differentiate control samples from dosed samples, the data were very easy to interpret, and signals from xenobiotic-related substances were easily removed without affecting the endogenous compounds. The proposed method is a complement or an alternative to NMR for metabolomic applications.

Use of control sample for estimation of prediction error in multivariate determination of lidocaine solutions with non-column chromatographic diode array UV spectroscopy.

The aim of this study was to investigate the ability of a control sample, of known content and identity, to diagnose and correct errors in the predictions when the same multivariate calibration model was used for analysis of new samples over time. A calibration set consisting of 16 samples with a known content of lidocaine was analysed and two external test sets, A and B, were used for the validation. Test set A contained 15 samples with different concentrations of lidocaine and test set B contained three samples with different lidocaine content, which were analysed six times in order to obtain a measure of repeatability. The multivariate calibration was done with PLS regression on UV spectra collected between 245 and 290 nm. A representative UV spectrum was exported from the collected DAD files by two methods, average spectrum over the whole file and average spectrum over the sample plug. Test set A was analysed further on another three occasions together with a control sample. The results showed that the control sample could be used to give a diagnosis and estimate of the prediction error. Moreover, the measured prediction error of the control sample could also be used to correct the predictions, thereby reducing the prediction error. Finally, some practical considerations regarding use of the proposed DAD method with a control sample are presented. The procedure suggested could lead to an efficient analytical approach where the same calibration model could be used over time without recalibration, which may be attractive in industrial quality control or screening analysis in pharmaceutical research.

Miniaturized on-line proteolysis-capillary liquid chromatography-mass spectrometry for peptide mapping of lactate dehydrogenase.

In this study, methodology was developed for on-line and miniaturized enzymatic digestion with liquid chromatographic (LC) separation and mass spectrometric (MS) detection. A packed capillary LC-MS system was combined with on-line trypsin cleavage of a model protein, lactate dehydrogenase, to provide an efficient system for peptide mapping. The protein was injected onto an enzymatic capillary reactor and the resulting peptides were efficiently trapped on a capillary trapping column. Different trapping columns were evaluated to achieve a high binding capacity for the peptides generated in the enzyme reactor. The peptides were further eluted from the pre-column and separated on an analytical capillary column by a buffer more suitable for the following an electrospray ionisation (ESI) MS process. An important aspect of the on-line approach was the desalting of peptides performed in the trapping column to avoid detrimental signal suppression in the ESI process. The developed on-line system was finally compared to a classical digestion in solution, with reference to peptide sequence coverage and sensitivity. It was shown that the on-line system gave more than 100% higher peptide sequence coverage than traditional digestion methods.