Decision threshold adjustment in class prediction

Standard classification algorithms are generally designed to maximize the number of correct predictions (concordance). The criterion of maximizing the concordance may not be appropriate in certain applications. In practice, some applications may emphasize high sensitivity (e.g., clinical diagnostic tests) and others may emphasize high specificity (e.g., epidemiology screening studies). This paper considers effects of the decision threshold on sensitivity, specificity, and concordance for four classification methods: logistic regression, classification tree, Fisher's linear discriminant analysis, and a weighted k-nearest neighbor. We investigated the use of decision threshold adjustment to improve performance of either sensitivity or specificity of a classifier under specific conditions. We conducted a Monte Carlo simulation showing that as the decision threshold increases, the sensitivity decreases and the specificity increases; but, the concordance values in an interval around the maximum concordance are similar. For specified sensitivity and specificity levels, an optimal decision threshold might be determined in an interval around the maximum concordance that meets the specified requirement. Three example data sets were analyzed for illustrations.     

Bioanalytical applications of aptamer and molecular-beacon probes in fluorescence-affinity assays

This critical review summarizes recent developments in highly sensitive, specific assays using nucleic-acid (NA)-affinity probes and fluorescence detection. We emphasize two groups of DNA and RNA probes (i.e. aptamers and molecular beacons) because of the increase in their bioanalytical applications. The affinity and the specificity of these NA probes combined with the diverse detection capability of fluorescence measurements (e.g., intensity, polarization, resonance-energy transfer and decay life-time) enable ultrasensitive assays for proteins, gene mutations, single-nucleotide polymorphisms and molecular-binding events.     

Molecular imprinting: a dynamic technique for diverse applications in analytical chemistry

Continuous advances in analyzing complex matrices, improving reliability and simplicity, and performing multiple simultaneous assays with extreme sensitivity are increasing. Several techniques have been developed for the quantitative assays of analytes at low concentrations (e.g., high-pressure liquid chromatography, gas chromatography, immunoassay and the polymerase chain reaction technique). To achieve highly specific and sensitive analysis, high affinity, stable, and specific recognition agents are needed. Although biological recognition agents are very specific and sensitive they are labile and/or have a low density of binding sites. During the past decade molecular imprinting has emerged as an attractive and highly accepted tool for the development of artificial recognition agents. Molecular imprinting is achieved by the interaction, either noncovalent or covalent, between complementary groups in a template molecule and functional monomer units through polymerization or polycondensation. These molecularly imprinted polymers have been widely employed for diverse applications (e.g., in chromatographic separation, drug screening, chemosensors, catalysis, immunoassays etc.) owing to their specificity towards the target molecules and high stability against physicochemical perturbations. In this review the advantages, applications, and recent developments in molecular imprinting technology are highlighted.     

Classification ensembles for unbalanced class sizes in predictive toxicology

This paper investigates the effects of the ratio of positive-to-negative samples on the sensitivity, specificity, and concordance. When the class sizes in the training samples are not equal, the classification rule derived will favor the majority class and result in a low sensitivity on the minority class prediction. We propose an ensemble classification approach to adjust for differential class sizes in a binary classifier system. An ensemble classifier consists of a set of base classifiers; its prediction rule is based on a summary measure of individual classifications by the base classifiers. Two re-sampling methods, augmentation and abatement, are proposed to generate different bootstrap samples of equal class size to build the base classifiers. The augmentation method balances the two class sizes by bootstrapping additional samples from the minority class, whereas the abatement method balances the two class sizes by sampling only a subset of samples from the majority class. The proposed procedure is applied to a data set to predict estrogen receptor binding activity and to a data set to predict animal liver carcinogenicity using SAR (structure-activity relationship) models as base classifiers. The abatement method appears to perform well in balancing sensitivity and specificity.     

Classification ensembles for unbalanced class sizes in predictive toxicology

This paper investigates the effects of the ratio of positive-to-negative samples on the sensitivity, specificity, and concordance. When the class sizes in the training samples are not equal, the classification rule derived will favor the majority class and result in a low sensitivity on the minority class prediction. We propose an ensemble classification approach to adjust for differential class sizes in a binary classifier system. An ensemble classifier consists of a set of base classifiers; its prediction rule is based on a summary measure of individual classifications by the base classifiers. Two re-sampling methods, augmentation and abatement, are proposed to generate different bootstrap samples of equal class size to build the base classifiers. The augmentation method balances the two class sizes by bootstrapping additional samples from the minority class, whereas the abatement method balances the two class sizes by sampling only a subset of samples from the majority class. The proposed procedure is applied to a data set to predict estrogen receptor binding activity and to a data set to predict animal liver carcinogenicity using SAR (structure-activity relationship) models as base classifiers. The abatement method appears to perform well in balancing sensitivity and specificity.     

Multi-residue screening of bovine urine on xenobiotic oestrogens with an oestrogen radioreceptor assay

An improved radioreceptor assay (RRA) is used for the screening of urine samples from cattle for the presence of exogenous oestrogenic anabolic compounds, e.g., stilbenes and zeranol. The method includes extraction of the hormones from urine samples with simultaneous purification using reversed-phase C18 cartridges. High-performance liquid chromatography is used to isolate the anabolics from the naturally occurring oestrogens. Fractions containing the stilbenes and zeranol are collected and subsequently analysed using the RRA with the oestrogen receptor, isolated from immature calf uteri, as a binder and tritiated 17 beta-oestradiol as a tracer. Urine samples from untreated calves and cows and samples from calves treated with zeranol-trenbolon acetate, dienoestrol or hexoestrol or samples containing diethylstilboestrol were analysed with this RRA method. Sensitivity, specificity and predictive values were calculated at different classification levels (0.4, 0.5, 0.6 and 1.0 ng/ml 'apparent' oestradiol in urine). An optimum sensitivity (89%) with a maximum specificity (95%) was reached at a classification level of 0.6 ng/ml. At this level the detection limits in urine samples are 0.5 ng/ml for hexoestrol, 0.6 ng/ml for diethylstilboestrol, 0.9 ng/ml for dienoestrol and 5.0 ng/ml for zeranol.     

Meta-analysis of gene expression for development and validation of a diagnostic biomarker panel for Oral Squamous Cell Carcinoma

We use a newly developed feature extraction and classification method to analyze previously published gene expression data sets in Oral Squamous Cell Carcinoma and in healthy oral mucosa in order to find a gene set sufficient for diagnoses. The feature selection technology is based on the relative dichotomy power concept published by us earlier. The resulting biomarker panel has 100% sensitivity and 95% specificity, is enriched in genes associated with oncogenesis and invasive tumor growth, and, unlike marker panels devised in earlier studies, shows concordance with previously published marker genes.     

High rate reversibility anode materials of lithium batteries from vapor-grown carbon nanofibers

We demonstrate the high-rate capability of lithium ion insertion-deinsertion reactions in carbon nanofibers (CNFs). The morphology of CNFs with structural and surface defects, due to the mixed features of disordered and graphitic carbon, plays an important role in both the enhancement of the lithium ion storage and the rate-determining reactions during the topotactic process. The reversible specific capacity of the CNFs at a 0.1 C rate was 461 mA x h/g. The most promising property, which is expected to overcome the hurdles of lithium batteries for high-power applications, is that they deliver considerably high specific capacity even at a very high charge-discharge current, i.e., at a cycling rate of 10 C the reversible capacity is around 170 mA x h/g with a 95% Coulombic efficiency.     

Pulse rise-time characterization of a high pressure xenon gamma detector for use in resolution enhancement

High pressure xenon ionization chamber detectors are possible alternatives to traditional thallium doped sodium iodide [NaI(Tl)] and hyperpure germanium as gamma-spectrometers in certain applications. Xenon detectors incorporating a Frisch grid exhibit energy resolutions comparable to cadmium/zinc/telluride (CZT) (e.g., 2% @ 662 keV) but with far greater sensitive volumes. The Frisch grid reduces the position dependence of the anode pulse rise-times, but it also increases the detector vibration sensitivity, anode capacitance, voltage requirements and mechanical complexity. We have been investigating the possibility of eliminating the grid electrode in high-pressure xenon detectors and preserving the high energy resolution using electronic rise-time compensation methods. A two-electrode cylindrical high pressure xenon gamma-detector coupled to time-to-amplitude conversion electronics was used to characterize the pulse rise-time of deposited gamma-photons. Time discrimination was used to characterize the pulse rise-time versus photo peak position and resolution. These data were collected to investigate the effect of pulse rise-time compensation on resolution and efficiency.     

High amplification rates from the association of two enzymes confined within a nanometric layer immobilized on an electrode: modeling and illustrating example

Electrochemical responses (e.g., chronoamperometric) obtained with an immobilized enzyme that produces an electroactive species may be used to quantitate the amount of enzyme or the concentration of its substrate. It is shown, on theoretical and experimental bases, that product-to-substrate coupling with a second enzyme co-immobilized with the first within one or within a small number of monolayers, allows high amplification rates (higher than 1000), avoids membrane transport limitations, and lends itself to precise kinetic analyses that provide guidelines for optimization of the analytical sensitivity. Very large amplification factors, as large as several thousands, can be reached experimentally, in agreement with appropriately derived theoretical predictions, thus opening the route to the rational design of high-performance substrate sensing or affinity assays applications.     

A medium for the presumptive detection of Enterobacter sakazakii in infant formula: interlaboratory study

A standard method for the detection of Enterobacteriaceae was modified for the presumptive detection of Enterobacter sakazakii, and the modified method was validated in an interlaboratory trial with 16 laboratories from 8 European countries. The modification included a differential-elective medium for the isolation of E. sakazakii, consisting of nutrient agar (NA) supplemented with 4-methyl-umbelliferyl alpha-D-glucoside (alpha-MUG). A 25 g sample was added to 225 mL buffered peptone water. After incubation at 35 degrees or 37 degrees C for 16 or 20 h, 10 mL nonselective enrichment was transferred into 90 mL selective enrichment. The selective enrichment was streaked on violet-red bile glucose agar (VRBGA) and incubated at 37 degrees C for 24 h. It was streaked in parallel on NA plates supplemented with alpha-MUG at 50 mg/L and incubated at 25 degrees C for 16 h, and afterwards for an additional 24 h at room temperature in the dark. E. sakazakii appeared as vivid yellow colonies under normal light and showed blue/violet fluorescence under UV light on NA + alpha-MUG plates. Validation samples represented powdered infant formula without E. sakazakii (blanks) and with low (1-10 colony-forming units [CFU]/25 g) and medium (1-10 CFU/g) contamination levels. All samples contained Pseudomonas aeruginosa and Lactobacillus spp. as background flora. The specificity for blank samples was 100%. The sensitivity of the low contamination level was similar for VRBGA and NA + alpha-MUG, i.e., 66.7% (66.7% accordance, 53.9% concordance). For the medium level the sensitivities were 96.7% (93.3% accordance, 93.5% concordance) for VRBGA and 98.3% (96.9% accordance, 96.9% concordance) for NA + alpha-MUG.     

Computer simulation study of probe-target hybridization in model DNA microarrays: effect of probe surface density and target concentration

We use lattice Monte Carlo simulations to study the thermodynamics of hybridization of single-stranded "target" genes in solution with complementary "probe" DNA molecules immobilized on a microarray surface. The target molecules in our system contain 48 segments and the probes tethered on a hard surface contain 8-24 segments. The segments on the probe and target are distinct, with each segment representing a sequence of nucleotides that interacts exclusively with its unique complementary target segment with a single hybridization energy; all other interactions are zero. We examine how surface density (number of probes per unit surface area) and concentration of target molecules affect the extent of hybridization. For short probe lengths, as the surface density increases, the probability of binding long stretches of target segments increases at low surface density, reaches a maximum at an intermediate surface density, and then decreases at high surface density. Furthermore, as the surface density increases, the target is less likely to bind completely to one probe; instead, it binds simultaneously to multiple probes. At short probe lengths, as the target concentration increases, the fraction of targets binding completely to the probes (specificity) decreases. At long probe lengths, varying the target concentration does not affect the specificity. At all target concentrations as the probe length increases, the fraction of target molecules bound to the probes by at least one segment (sensitivity) increases while the fraction of target molecules completely bound to the probes (specificity) decreases. This work provides general guidelines to maximizing microarray sensitivity and specificity. Our results suggest that the sensitivity and specificity can be maximized by using probes 130-180 nucleotides long at a surface density in the range of 7 x 10(-5)- 3 x 10(-4) probe molecules per nm(2).     

Ultralow flow liquid chromatography and related approaches: A focus on recent bioanalytical applications

Ultralow flow LC employs ultra-narrow bore columns and mid-range pL/min to low nL/min flow rates (i.e., ≤20 nL/min). The separation columns that are used under these conditions are typically 2–30 μm in inner diameter. Ultralow flow LC systems allow for exceptionally high sensitivity and frequently high resolution. There has been an increasing interest in the analysis of scarce biological samples, for example, circulating tumor cells, extracellular vesicles, organelles, and single cells, and ultralow flow LC was efficiently applied to such samples. Hence, advances towards dedicated ultralow flow LC instrumentation, technical approaches, and higher throughput (e.g., tens-to-hundreds of single cells analyzed per day) were recently made. Here, we review the types of ultralow flow LC technology, followed by a discussion of selected representative ultralow flow LC applications, focusing on the progress made in bioanalysis of amount-limited samples during the last 10 years. We also discuss several recently reported high-sensitivity applications utilizing flow rates up to 100 nL/min, which are below commonly used nanoLC flow rates. Finally, we discuss the path forward for future developments of ultralow flow LC.     

Simultaneous determination of nine intense sweeteners in foodstuffs by high performance liquid chromatography and evaporative light scattering detection--development and single-laboratory validation

A high performance liquid chromatographic method with evaporative light scattering detection (HPLC-ELSD) has been developed for the simultaneous determination of multiple sweeteners, i.e., acesulfame-K, alitame, aspartame, cyclamic acid, dulcin, neotame, neohesperidine dihydrochalcone, saccharin and sucralose in carbonated and non-carbonated soft drinks, canned or bottled fruits and yoghurt. The procedure involves an extraction of the nine sweeteners with a buffer solution, sample clean-up using solid-phase extraction cartridges followed by an HPLC-ELSD analysis. The trueness of the method was satisfactory with recoveries ranging from 93 to 109% for concentration levels around the maximum usable dosages for authorised sweeteners and from 100 to 112% for unauthorised compounds at concentration levels close to the limit of quantification (LOQs). Precision measures showed mean repeatability values of <4% (expressed as relative standard deviation) for highly concentrated samples and <5% at concentration levels close to the LOQs. Intermediate precision was in most cases <8%. The limits of detection (LODs) were below 15 microg g(-1) and the LOQs below 30 microg g(-1) in all three matrices. Only dulcin showed slightly higher values, i.e., LODs around 30 microg g(-1) and LOQs around 50 microg g(-1)     

Matching the laser wavelength to the absorption properties of matrices increases the ion yield in UV-MALDI mass spectrometry

A high analytical sensitivity in ultraviolet matrix-assisted laser desorption ionization mass spectrometry (MALDI-MS) is only achieved if the laser wavelength corresponds to a high optical absorption of the matrix. Laser fluence and the physicochemical properties of the compounds, e.g., the proton affinity, also influence analytical sensitivity significantly. In combination, these parameters determine the amount of material ejected per laser pulse and the ion yield, i.e., the fraction of ionized biomolecules. Here, we recorded peptide ion signal intensities as a function of these parameters. Three cinnamic acid matrices were investigated: α-cyano-4-hydroxycinnamic acid, α-cyano-4-chlorocinnamic acid, and α-cyano-2,4-difluorocinnamic acid. In addition, 2,5-dihydroxybenzoic acid was used in comparison experiments. Ion signal intensities “per laser shot” and integrated ion signal intensities were acquired over 900 consecutive laser pulses applied on distinct positions on the dried-droplet sample preparations. With respect to laser wavelength, the two standard MALDI wavelengths of 337/355 nm were investigated. Also, 305 or 320 nm was selected to account for the blue-shifted absorption profiles of the halogenated derivatives. Maximal peptide ion intensities were obtained if the laser wavelength fell within the peak of the absorption profile of the compound and for fluences two to three times the corresponding ion detection threshold. The results indicate ways for improving the analytical sensitivity in MALDI-MS, and in particular for MALDI-MS imaging applications where a limited amount of material is available per irradiated pixel.     

Recent developments in radiometric and mass spectrometry methods for marine radioactivity measurements

The most important recent developments in radiometric techniques have been the operation of high efficiency HPGe detectors with anticosmic or antiCompton shielding often placed underground, ship-board measurements of ²³⁴Th using gamma-spectrometry or beta-counting and underwater gamma-spectrometry. In mass spectrometry techniques, the availability of high resolution ICP-MS and applications of AMS for the analysis of long-lived radionuclides have opened doors for investigations which previously required too large samples, or were not possible because of lack of sensitivity. These developments are illustrated by several examples of marine radioactivity studies, which include contaminated sites (e.g., Fangataufa Atoll) and open ocean studies (e.g., the Atlantic Ocean) as well as analyses of IAEA reference materials. A comparison of Pu results obtained by alpha-spectrometry, ICP-MS and AMS has shown that a reasonably good agreement has been obtained between all three methods.     

Merits of random forests emerge in evaluation of chemometric classifiers by external validation

Real-world applications will inevitably entail divergence between samples on which chemometric classifiers are trained and the unknowns requiring classification. This has long been recognized, but there is a shortage of empirical studies on which classifiers perform best in ‘external validation’ (EV), where the unknown samples are subject to sources of variation relative to the population used to train the classifier. Survey of 286 classification studies in analytical chemistry found only 6.6% that stated elements of variance between training and test samples. Instead, most tested classifiers using hold-outs or resampling (usually cross-validation) from the same population used in training. The present study evaluated a wide range of classifiers on NMR and mass spectra of plant and food materials, from four projects with different data properties (e.g., different numbers and prevalence of classes) and classification objectives. Use of cross-validation was found to be optimistic relative to EV on samples of different provenance to the training set (e.g., different genotypes, different growth conditions, different seasons of crop harvest). For classifier evaluations across the diverse tasks, we used ranks-based non-parametric comparisons, and permutation-based significance tests. Although latent variable methods (e.g., PLSDA) were used in 64% of the surveyed papers, they were among the less successful classifiers in EV, and orthogonal signal correction was counterproductive. Instead, the best EV performances were obtained with machine learning schemes that coped with the high dimensionality (914–1898 features). Random forests confirmed their resilience to high dimensionality, as best overall performers on the full data, despite being used in only 4.5% of the surveyed papers. Most other machine learning classifiers were improved by a feature selection filter (ReliefF), but still did not out-perform random forests.     

Implementation of quantum gate operations in molecules with weak laser fields

We numerically propose a way to perform quantum computations by combining an ensemble of molecular states and weak laser pulses. A logical input state is expressed as a superposition state (a wave packet) of molecular states, which is initially prepared by a designed femtosecond laser pulse. The free propagation of the wave packet for a specified time interval leads to the specified change in the relative phases among the molecular basis states, which corresponds to a computational result. The computational results are retrieved by means of quantum interferometry. Numerical tests are implemented in the vibrational states of the B state of I2 employing controlled-NOT gate, and 2 and 3 qubits Fourier transforms. All the steps involved in the computational scheme, i.e., the initial preparation, gate operation, and detection steps, are achieved with extremely high precision.     

Ordered porous thin films in electrochemical analysis

Recently, the field of highly-ordered mesoporous and macroporous thin films coated onto solid electrode surfaces has begun to receive attention due to their great interest for electrochemical analysis. This review highlights the features of both electricallyconducting and non-conducting organized layers, which are applicable to designing electrochemical sensors, and the methods applied to construct these novel nanomaterials.We emphasize methods based on use of self-assembled colloidal templates (e.g., surfactants or nanoparticles), around which the materials of interest are formed. We then describe their basic electrochemical behavior and discuss their possible use as electrochemical sensors and biosensors, mostly in the particular case of structured metallic layers, functionalized mesoporous silica films, and some other continuous three-dimensional ordered porous structures.     

Double pulse laser ablation and plasma: Laser induced breakdown spectroscopy signal enhancement

A review of recent results of the studies of double laser pulse plasma and ablation for laser induced breakdown spectroscopy applications is presented. The double pulse laser induced breakdown spectroscopy configuration was suggested with the aim of overcoming the sensitivity shortcomings of the conventional single pulse laser induced breakdown spectroscopy technique. Several configurations have been suggested for the realization of the double pulse laser induced breakdown spectroscopy technique: collinear, orthogonal pre-spark, orthogonal pre-heating and dual pulse crossed beam modes. In addition, combinations of laser pulses with different wavelengths, different energies and durations were studied, thus providing flexibility in the choice of wavelength, pulse width, energy and pulse sequence. The double pulse laser induced breakdown spectroscopy approach provides a significant enhancement in the intensity of laser induced breakdown spectroscopy emission lines up to two orders of magnitude greater than a conventional single pulse laser induced breakdown spectroscopy. The double pulse technique leads to a better coupling of the laser beam with the plasma plume and target material, thus providing a more temporally effective energy delivery to the plasma and target. The experimental results demonstrate that the maximum effect is obtained at some optimum separation delay time between pulses. The optimum value of the interpulse delay depends on several factors, such as the target material, the energy level of excited states responsible for the emission, and the type of enhancement process considered. Depending on the specified parameter, the enhancement effects were observed on different time scales ranging from the picosecond time level (e.g., ion yield, ablation mass) up to the hundred microsecond level (e.g., increased emission intensity for laser induced breakdown spectroscopy of submerged metal target in water). Several suggestions have been proposed to explain the mechanism of double pulse enhancement.