On the optimal usage of labelled examples in semi-supervised multi-class classification problems

In recent years, the performance of semi-supervised learning has been theoretically investigated. However, most of this theoretical development has focussed on binary classification problems. In this paper, we take it a step further by extending the work of Castelli and Cover [1] [2] to the multi-class paradigm. Particularly, we consider the key problem in semi-supervised learning of classifying an unseen instance x into one of K different classes, using a training dataset sampled from a mixture density distribution and composed of l labelled records and u unlabelled examples. Even under the assumption of identifiability of the mixture and having infinite unlabelled examples, labelled records are needed to determine the K decision regions. Therefore, in this paper, we first investigate the minimum number of labelled examples needed to accomplish that task. Then, we propose an optimal multi-class learning algorithm which is a generalisation of the optimal procedure proposed in the literature for binary problems. Finally, we make use of this generalisation to study the probability of error when the binary class constraint is relaxed.

A spectroscopic investigation of pigment and ceramic samples from Copán, Honduras

A combination of micro-Raman spectroscopy, micro-infrared spectroscopy and SEM–EDX was employed to characterize decorative pigments on Classic Maya ceramics from Copán, Honduras. Variation in red paint mixtures was correlated with changing ceramic types and improvements in process and firing techniques. We have confirmed the use of specular hematite on Coner ceramics by the difference in intensities of Raman bands. Different compositions of brown paint were correlated with imported and local wares. The carbon-iron composition of the ceramic type, Surlo Brown, was confirmed. By combining micro-Raman analysis with micro-ATR infrared and SEM–EDX, we have achieved a more comprehensive characterization of the paint mixtures. These spectroscopic techniques can be used non-destructively on raw samples as a rapid confirmation of ceramic type.

Modification and validation of a microwave-assisted digestion method for subsequent ICP-MS determination of selected heavy metals in sediment and fish samples in Agusan River, Philippines

This study investigated, validated, and applied the optimum conditions for a modified microwave assisted digestion method for subsequent ICP-MS determination of mercury, cadmium, and lead in two matrices relevant to water quality, that is, sediment and fish. Three different combinations of power, pressure, and time conditions for microwave-assisted digestion were tested, using two certified reference materials representing the two matrices, to determine the optimum set of conditions. Validation of the optimized method indicated better recovery of the studied metals compared to standard methods. The validated method was applied to sediment and fish samples collected from Agusan River and one of its tributaries, located in Eastern Mindanao, Philippines. The metal concentrations in sediment ranged from 2.85 to 341.06 mg/kg for Hg, 0.05 to 44.46 mg/kg for Cd and 2.20 to 1256.16 mg/kg for Pb. The results indicate that the concentrations of these metals in the sediments rapidly decrease with distance downstream from sites of contamination. In the selected fish species, the metals were detected but at levels that are considered safe for human consumption, with concentrations of 2.14 to 6.82 μg/kg for Hg, 0.035 to 0.068 μg/kg for Cd, and 0.019 to 0.529 μg/kg for Pb.

Simultaneous voltammetric determination of three herbicides in food and water samples with the aid of chemometrics

Differential pulse stripping voltammetry method(DPSV) was applied to the determination of three herbicides, ametryn, cyanatryn, and dimethametryn. It was found that their voltammograms overlapped strongly, and it is difficult to determine these compounds individually from their mixtures. With the aid of chemometrics, classical least squares(CLS), principal component regression(PCR) and partial least squares(PLS), voltammogram resolution and quantitative analysis of the synthetic mixtures of the three compounds were successfully performed. The proposed method was also applied to the analysis of some real samples with satisfactory results.

Sequential fusion of decisions from adaptive and random samples for controlled verification of errors

Fusion techniques have received considerable attention for achieving performance improvement with biometrics. While a multi-sample fusion architecture reduces false rejects, it also increases false accepts. This impact on performance also depends on the nature of subsequent attempts, i.e., random or adaptive. Expressions for error rates are presented and experimentally evaluated in this work by considering the multi-sample fusion architecture for text-dependent speaker verification using HMM based digit dependent speaker models. Analysis incorporating correlation modeling demonstrates that the use of adaptive samples improves overall fusion performance compared to randomly repeated samples. For a text dependent speaker verification system using digit strings, sequential decision fusion of seven instances with three random samples is shown to reduce the overall error of the verification system by 26% which can be further reduced by 6% for adaptive samples. This analysis novel in its treatment of random and adaptive multiple presentations within a sequential fused decision architecture, is also applicable to other biometric modalities such as finger prints and handwriting samples.

Use of serum and blood samples on filter paper to improve the surveillance of dengue in Pacific Island Countries

Background In Pacific Island Countries (PICs) the epidemiology of dengue is characterized by long-term transmission of a single dengue virus (DENV) serotype. The emergence of a new serotype in one island country often indicates major outbreaks with this serotype will follow in other PICs. Objectives Filter paper (FP) cards on which whole blood or serum from dengue suspected patients had been dried was evaluated as a method for transportation of this material by standard mail delivery throughout the Pacific. Study design Twenty-two FP-dried whole blood samples collected from patients in New Caledonia and Wallis & Futuna Islands, during DENV-1 and DENV-4 transmission, and 76 FP-dried sera collected from patients in Yap State, Majuro (Republic of Marshall Islands), Tonga and Fiji, before and during outbreaks of DENV-2 in Yap State and DENV-4 in Majuro, were tested for the presence of DENV RNA, by serotype specific RT-PCR, at the Institut Louis Malardé in French Polynesia. Results The serotype of DENV could be determined, by a variety of RT-PCR procedures, in the FP-dried samples after more than three weeks of transport at ambient temperatures. In most cases, the sequencing of the envelope gene to genotype the viruses also was possible. Conclusions The serotype and genotype of DENV can be determined from FP-dried serum or whole blood samples transported over thousands of kilometers at ambient, tropical, temperatures. This simple and low-cost approach to virus identification should be evaluated in isolated and resource poor settings for surveillance for a range of significant viral diseases.

Absolute quantitation of Marek's disease virus and Herpesvirus of turkeys in chicken lymphocyte, feather tip and dust samples using real-time PCR

The further development of Taqman quantitative real-time PCR (qPCR) assays for the absolute quantitation of Marek's disease virus serotype 1 (MDV1) and Herpesvirus of turkeys (HVT) viruses is described and the sensitivity and reproducibility of each assay reported. Using plasmid DNA copies, the lower limit of detection was determined to be 5 copies for the MDV1 assay and 75 copies for the HVT assay. Both assays were found to be highly reproducible for Ct values and calculated copy numbers with mean intra- and inter-assay coefficients of variation being less than 5% for Ct and 20% for calculated copy number. The genome copy number of MDV1 and HVT viruses was quantified in PBL and feather tips from experimentally infected chickens, and field poultry dust samples. Parallelism was demonstrated between the plasmid-based standard curves, and standard curves derived from infected spleen material containing both viral and host DNA, allowing the latter to be used for absolute quantification. These methods should prove useful for the reliable differentiation and absolute quantitation of MDV1 and HVT viruses in a wide range of samples.

Association study of the dystrobrevin-binding gene with schizophrenia in Australian and Indian samples

Numerous studies have reported association between variants in the dystrobrevin binding protein 1 (dysbindin) gene (DTNBP1) and schizophrenia. However, the pattern of results is complex and to date, no specific risk marker or haplotype has been consistently identified. The number of single nucleotide polymorphisms (SNPs) tested in these studies has ranged from 5 to 20. We attempted to replicate previous findings by testing 16 SNPs in samples of 41 Australian pedigrees, 194 Australian cases and 180 controls, and 197 Indian pedigrees. No globally significant evidence for association was observed in any sample, despite power calculations indicating sufficient power to replicate several previous findings. Possible explanations for our results include sample differences in background linkage disequilibrium and/or risk allele effect size, the presence of multiple risk alleles upon different haplotypes, or the presence of a single risk allele upon multiple haplotypes. Some previous associations may also represent false positives. Examination of Caucasian HapMap phase II genotype data spanning the DTNBP1 region indicates upwards of 40 SNPs are required to satisfactorily assess all nonredundant variation within DTNBP1 and its potential regulatory regions for association with schizophrenia. More comprehensive studies in multiple samples will be required to determine whether specific DTNBP1 variants function as risk factors for schizophrenia.

Critical review on the stability of illicit drugs in sewers and wastewater samples

Wastewater-based epidemiology (WBE) applies advanced analytical methods to quantify drug residues in wastewater with the aim to estimate illicit drug use at the population level. Transformation processes during transport in sewers (chemical and biological reactors) and storage of wastewater samples before analysis are expected to change concentrations of different drugs to varying degrees. Ignoring transformation for drugs with low to medium stability will lead to an unknown degree of systematic under- or overestimation of drug use, which should be avoided. This review aims to summarize the current knowledge related to the stability of commonly investigated drugs and, furthermore, suggest a more effective approach to future experiments. From over 100 WBE studies, around 50 mentioned the importance of stability and 24 included tests in wastewater. Most focused on in-sample stability (i.e., sample preparation, preservation and storage) and some extrapolated to in-sewer stability (i.e., during transport in real sewers). While consistent results were reported for rather stable compounds (e.g., MDMA and methamphetamine), a varying range of stability under different or similar conditions was observed for other compounds (e.g., cocaine, amphetamine and morphine). Wastewater composition can vary considerably over time, and different conditions prevail in different sewer systems. In summary, this indicates that more systematic studies are needed to: i) cover the range of possible conditions in sewers and ii) compare results more objectively. To facilitate the latter, we propose a set of parameters that should be reported for in-sewer stability experiments. Finally, a best practice of sample collection, preservation, and preparation before analysis is suggested in order to minimize transformation during these steps.

Abnormal levels of heterogeneous nuclear ribonucleoprotein A2B1 (hnRNPA2B1) in tumour tissue and blood samples from patients diagnosed with lung cancer

Lung cancer is the second most common type of cancer in the world and is the most common cause of cancer-related death in both men and women. Research into causes, prevention and treatment of lung cancer is ongoing and much progress has been made recently in these areas, however survival rates have not significantly improved. Therefore, it is essential to develop biomarkers for early diagnosis of lung cancer, prediction of metastasis and evaluation of treatment efficiency, as well as using these molecules to provide some understanding about tumour biology and translate highly promising findings in basic science research to clinical application. In this investigation, two-dimensional difference gel electrophoresis and mass spectrometry were initially used to analyse conditioned media from a panel of lung cancer and normal bronchial epithelial cell lines. Significant proteins were identified with heterogeneous nuclear ribonucleoprotein A2B1 (hnRNPA2B1), pyruvate kinase M2 isoform (PKM2), Hsc-70 interacting protein and lactate dehydrogenase A (LDHA) selected for analysis in serum from healthy individuals and lung cancer patients. hnRNPA2B1, PKM2 and LDHA were found to be statistically significant in all comparisons. Tissue analysis and knockdown of hnRNPA2B1 using siRNA subsequently demonstrated both the overexpression and potential role for this molecule in lung tumorigenesis. The data presented highlights a number of in vitro derived candidate biomarkers subsequently verified in patient samples and also provides some insight into their roles in the complex intracellular mechanisms associated with tumour progression.