Supervised learning with decision margins in pools of spiking neurons

Learning to categorise sensory inputs by generalising from a few examples whose category is precisely known is a crucial step for the brain to produce appropriate behavioural responses. At the neuronal level, this may be performed by adaptation of synaptic weights under the influence of a training signal, in order to group spiking patterns impinging on the neuron. Here we describe a framework that allows spiking neurons to perform such “supervised learning”, using principles similar to the Support Vector Machine, a well-established and robust classifier. Using a hinge-loss error function, we show that requesting a margin similar to that of the SVM improves performance on linearly non-separable problems. Moreover, we show that using pools of neurons to discriminate categories can also increase the performance by sharing the load among neurons.     

Growth, seed development and genetic analysis in wild type and Def mutant of Pisum sativum L

Background

Dementia and cognitive impairment associated with aging are a major medical and social concern. Neuropsychological testing is a key element in the diagnostic procedures of Mild Cognitive Impairment (MCI), but has presently a limited value in the prediction of progression to dementia. We advance the hypothesis that newer statistical classification methods derived from data mining and machine learning methods like Neural Networks, Support Vector Machines and Random Forests can improve accuracy, sensitivity and specificity of predictions obtained from neuropsychological testing. Seven non parametric classifiers derived from data mining methods (Multilayer Perceptrons Neural Networks, Radial Basis Function Neural Networks, Support Vector Machines, CART, CHAID and QUEST Classification Trees and Random Forests) were compared to three traditional classifiers (Linear Discriminant Analysis, Quadratic Discriminant Analysis and Logistic Regression) in terms of overall classification accuracy, specificity, sensitivity, Area under the ROC curve and Press'Q. Model predictors were 10 neuropsychological tests currently used in the diagnosis of dementia. Statistical distributions of classification parameters obtained from a 5-fold cross-validation were compared using the Friedman's nonparametric test.

Results

Press' Q test showed that all classifiers performed better than chance alone (p < 0.05). Support Vector Machines showed the larger overall classification accuracy (Median (Me) = 0.76) an area under the ROC (Me = 0.90). However this method showed high specificity (Me = 1.0) but low sensitivity (Me = 0.3). Random Forest ranked second in overall accuracy (Me = 0.73) with high area under the ROC (Me = 0.73) specificity (Me = 0.73) and sensitivity (Me = 0.64). Linear Discriminant Analysis also showed acceptable overall accuracy (Me = 0.66), with acceptable area under the ROC (Me = 0.72) specificity (Me = 0.66) and sensitivity (Me = 0.64). The remaining classifiers showed overall classification accuracy above a median value of 0.63, but for most sensitivity was around or even lower than a median value of 0.5.

Conclusions

When taking into account sensitivity, specificity and overall classification accuracy Random Forests and Linear Discriminant analysis rank first among all the classifiers tested in prediction of dementia using several neuropsychological tests. These methods may be used to improve accuracy, sensitivity and specificity of Dementia predictions from neuropsychological testing.     

Hadamard Kernel SVM with applications for breast cancer outcome predictions

Background

Breast cancer is one of the leading causes of deaths for women. It is of great necessity to develop effective methods for breast cancer detection and diagnosis. Recent studies have focused on gene-based signatures for outcome predictions. Kernel SVM for its discriminative power in dealing with small sample pattern recognition problems has attracted a lot attention. But how to select or construct an appropriate kernel for a specified problem still needs further investigation.

Results

Here we propose a novel kernel (Hadamard Kernel) in conjunction with Support Vector Machines (SVMs) to address the problem of breast cancer outcome prediction using gene expression data. Hadamard Kernel outperform the classical kernels and correlation kernel in terms of Area under the ROC Curve (AUC) values where a number of real-world data sets are adopted to test the performance of different methods.

Conclusions

Hadamard Kernel SVM is effective for breast cancer predictions, either in terms of prognosis or diagnosis. It may benefit patients by guiding therapeutic options. Apart from that, it would be a valuable addition to the current SVM kernel families. We hope it will contribute to the wider biology and related communities.

     

Using multitask classification methods to investigate the kinase-specific phosphorylation sites

Background

Identification of phosphorylation sites by computational methods is becoming increasingly important because it reduces labor-intensive and costly experiments and can improve our understanding of the common properties and underlying mechanisms of protein phosphorylation.

Methods

A multitask learning framework for learning four kinase families simultaneously, instead of studying each kinase family of phosphorylation sites separately, is presented in the study. The framework includes two multitask classification methods: the Multi-Task Least Squares Support Vector Machines (MTLS-SVMs) and the Multi-Task Feature Selection (MT-Feat3).

Results

Using the multitask learning framework, we successfully identify 18 common features shared by four kinase families of phosphorylation sites. The reliability of selected features is demonstrated by the consistent performance in two multi-task learning methods.

Conclusions

The selected features can be used to build efficient multitask classifiers with good performance, suggesting they are important to protein phosphorylation across 4 kinase families.

     

Fluorescence imaging of microbe-containing particles shot from a two-stage Light-gas gun into an aerogel

We have proposed an experiment (the Tanpopo mission) to capture microbes on the Japan Experimental Module of the International Space Station. An ultra low-density silica aerogel will be exposed to space for more than 1 year. After retrieving the aerogel, particle tracks and particles found in it will be visualized by fluorescence microscopy after staining it with a DNA-specific fluorescence dye. In preparation for this study, we simulated particle trapping in an aerogel so that methods could be developed to visualize the particles and their tracks. During the Tanpopo mission, particles that have an orbital velocity of ~8 km/s are expected to collide with the aerogel. To simulate these collisions, we shot Deinococcus radiodurans-containing Lucentite particles into the aerogel from a two-stage light-gas gun (acceleration 4.2 km/s). The shapes of the captured particles, and their tracks and entrance holes were recorded with a microscope/camera system for further analysis. The size distribution of the captured particles was smaller than the original distribution, suggesting that the particles had fragmented. We were able to distinguish between microbial DNA and inorganic compounds after staining the aerogel with the DNA-specific fluorescence dye SYBR green I as the fluorescence of the stained DNA and the autofluorescence of the inorganic particles decay at different rates. The developed methods are suitable to determine if microbes exist at the International Space Station altitude.     

Learning from positive examples when the negative class is undetermined- microRNA gene identification

Background

The application of machine learning to classification problems that depend only on positive examples is gaining attention in the computational biology community. We and others have described the use of two-class machine learning to identify novel miRNAs. These methods require the generation of an artificial negative class. However, designation of the negative class can be problematic and if it is not properly done can affect the performance of the classifier dramatically and/or yield a biased estimate of performance. We present a study using one-class machine learning for microRNA (miRNA) discovery and compare one-class to two-class approaches using naïve Bayes and Support Vector Machines. These results are compared to published two-class miRNA prediction approaches. We also examine the ability of the one-class and two-class techniques to identify miRNAs in newly sequenced species.

Results

Of all methods tested, we found that 2-class naive Bayes and Support Vector Machines gave the best accuracy using our selected features and optimally chosen negative examples. One class methods showed average accuracies of 70–80% versus 90% for the two 2-class methods on the same feature sets. However, some one-class methods outperform some recently published two-class approaches with different selected features. Using the EBV genome as and external validation of the method we found one-class machine learning to work as well as or better than a two-class approach in identifying true miRNAs as well as predicting new miRNAs.

Conclusion

One and two class methods can both give useful classification accuracies when the negative class is well characterized. The advantage of one class methods is that it eliminates guessing at the optimal features for the negative class when they are not well defined. In these cases one-class methods can be superior to two-class methods when the features which are chosen as representative of that positive class are well defined.

Availability

The OneClassmiRNA program is available at: [1]

     

Compound analysis via graph kernels incorporating chirality

High accuracy is paramount when predicting biochemical characteristics using Quantitative Structural-Property Relationships (QSPRs). Although existing graph-theoretic kernel methods combined with machine learning techniques are efficient for QSPR model construction, they cannot distinguish topologically identical chiral compounds which often exhibit different biological characteristics. In this paper, we propose a new method that extends the recently developed tree pattern graph kernel to accommodate stereoisomers. We show that Support Vector Regression (SVR) with a chiral graph kernel is useful for target property prediction by demonstrating its application to a set of human vitamin D receptor ligands currently under consideration for their potential anti-cancer effects.     

A bibliometric investigation of life cycle assessment research in the web of science databases

Purpose

Over the past few decades, life cycle assessment (LCA) methodologies have been developed extensively, and there has been a growing interest in LCA research. However, as attested by scientific literature, few systematic, synthesizing, and visualizing studies have been found on LCA research which show how this field has evolved over time. The goal of this mainly bibliometric, empirical study is to get insight into publication performance of global LCA research, characterize its intellectual structure, and trace its evolution by using the bibliometric method with visual mapping.

Methods

Based on the data from the ISI Web of Science databases Science Citation Index Expanded (SCI-EXPANDED), Social Sciences Citation Index (SSCI), Conference Proceedings Citation Index—Science (CPCI-S) and Conference Proceedings Citation Index —Social Science & Humanities (CPCI-SSH) in the period of 1998–2013, bibliometric methods are used to investigate general development profiles of LCA research, while knowledge domain visualization technologies are employed to conduct a further co-citation analysis.

Results and discussion

The results and discussions of this research mainly shed light on (1) basic statistics of significant publication performances, (2) research focuses and their intellectual base in LCA research, (3) how the streams of research evolved during the whole period of interest.

Conclusions

A new work on systematic and synthesizing study is conducted in this research to evaluate and map LCA research-related context. Some salient scholarly journals and institutions are identified that have shown a significant impact during the exponential growth of LCA research in the past 16 years. Biofuel, process design, solid waste management, and livestock production-related LCA researches are the main areas where interest is surging, confirmed by the active citers in each specialty. Furthermore, from the perspective of science mapping, evolution of LCA research is traced and some pivot publications are identified, which work as structural holes for the LCA-research development in the given time window.     

Adaptive mining prediction model for content recommendation to coronary heart disease patients

This paper proposes the Fuzzy Rule-based Adaptive Coronary Heart Disease Prediction Support Model (FbACHD_PSM), which gives content recommendation to coronary heart disease patients. The proposed model uses a mining technique validated by medical experts to provide recommendations. FbACHD_PSM consists of three parts for heart disease risk prediction. First, a fuzzy membership function is constructed using medical guidelines and statistical methods. Then, a decision-tree rule induction technique creates mining-based rules that are subjected to validation by medical experts. As the rules may not be medically suitable, the experts add rules that have been verified and delete inappropriate rules. Thirdly, using fuzzy inference based on Mamdani’s method, the model predicts the risk of heart disease. Based on this, final recommendations are provided to patients regarding normal living, nutrition control, exercise, and drugs. To implement our proposed model and evaluate its performance, we use a dataset from a single tertiary hospital.     

Lack of association between the G+2044A polymorphism of interleukin-13 gene and chronic obstructive pulmonary disease: a meta-analysis

Numerous studies have investigated association of interleukin-13 (IL-13) G+2044A polymorphism with COPD susceptibility; however, the results were inconsistent and inconclusive. To evaluate the association between the IL-13 G+2044A polymorphism and susceptibility to COPD, a meta-analysis of published case–control studies was performed. Based on PubMed and Chinese database, this research selected studies that examined the association of the IL-13 G+2044A polymorphism with COPD. A genetic model-free approach was used to assess whether the combined data showed this association. Then a subgroup analysis was also performed, with stratifications for race, study design, and sample size. Six studies (total 1,213 COPD patients and 801 control subjects) for the IL-13 G+2044A polymorphism with COPD were included in the meta-analysis (G- vs A-allele: OR 1.12, 95 % CI 0.96–1.32, P = 0.15; genotypes GG+GA vs genotype AA: OR 0.99, 95 % CI 0.49–2.00, P = 0.98; genotype GG vs genotypes GA+AA: OR 1.18, 95 % CI 0.97–1.44, P = 0.09; genotype GA vs genotypes GG+AA: OR 0.85, 95 % CI 0.70–1.04, P = 0.11). This meta-analysis demonstrates that the IL-13 G+2044A polymorphism does not confer susceptibility to COPD. More detailed data about individual and environment, larger sample sizes with unbiased genotyping methods and matched controls in different populations are required.     

Antagonistic interactions between Aureobasidium pullulans and Fusarium culmorum,a fungal pathogen of winter wheat

The effect of the biological control agent Aureobasidium pullulans (de Bary) G. Arnaud on the development of Fusarium head blight (FHB) on winter wheat and kernel contamination with fungi of the genera Fusarium, Acremonium, Cladosporium and Penicillium was analyzed in a greenhouse experiment. Scanning electron microscopy was used to evaluate the distribution of A. pullulans cells and aggregates on wheat kernels, infection structures of Fusarium culmorum (W.G. Smith) Sacc and the antagonist-pathogen interactions. Biological control with A. pullulans reduced FHB severity by 21.67 % and improved grain filling by 5.02 %, compared with the control treatment. The survival of A. pullulans was good (to 31 cells per kernel), in particular on the surface and in the crease of kernels, including in pathogen-inoculated wheat plants. A. pullulans cells firmly adhered to F. culmorum hyphae, and damaged them. In most cases, autochthonous communities of filamentous fungi of the genera Acremonium and Penicillium developed at a slower rate after kernel inoculation with the pathogen.     

Teaching Bernoulli's Principle through Demonstrations

ABSTRACT

One proven strategy to help students make sense of abstract concepts is to sequence instruction so students have exploratory opportunities to investigate science before being introduced to new science explanations (Abraham and Renner 1986 Abraham, M. R. and Renner, J. W. 1986. The sequence of learning cycle activities in high school chemistry. Journal of Research in Science Teaching, 23: 121–143. [Crossref], [Web of Science ®] , [Google Scholar]; Renner, Abraham, and Birnie 1988 Renner, J. W., Abraham, M. R. and Birnie, H. H. 1988. The necessity of each phase of the learning cycle in teaching high school physics. Journal of Research in Science Teaching, 25: 39–58. [Crossref], [Web of Science ®] , [Google Scholar]). To help physical science teachers make sense of how to effectively sequence lessons, this article summarizes our experiences using an exploration–explanation sequence of instruction to teach Bernoulli's principle to prospective middle and secondary science teachers in a science methods course. We use demonstrations during our Bernoulli unit to help students go back and forth between their observations of phenomenon and what occurs on the microscopic level with what we have termed molecular talk. Students engage in guiding questions, consider their old and new understandings of science, and use evidence to construct new ideas during all stages of the lesson.     

The localization of auxin transporters PIN3 and LAX3 during lateral root development in Arabidopsis thaliana

Fluorophore tagged proteins are used in Arabidopsis thaliana to understand their functional role in plant development. This requires the analysis of their spatial localization in planta. However, the localization analysis is often perturbed by a significant overlap of the fluorophores used to label proteins of interest and the optical filtering methods available on the confocal microscope. This problem can be addressed by the use of spectral imaging with linear unmixing the image data. We applied this method to help us identify double transgenic A. thaliana lines which expressed two fluorescently tagged auxin transporter proteins: the auxin efflux protein PIN-FORMED-3 (PIN3), tagged with green fluorescent protein (GFP), and the auxin influx protein LIKE-AUX1-3 (LAX3), tagged with yellow fluorescent protein (YFP). This method allows the reliable separation of overlapping GFP and YFP fluorescence signals and subsequent localization analysis highlighting the potential benefit of this methodology in studies of lateral root development.     

Different Biomarkers in Non-Small Cell Lung Cancer in Blood Vessel Invasion

In clinical settings, lung cancer is divided into small cell lung cancer and non-small cell lung cancer, and chemotherapy is depended on the difference. Using the same chemotherapy treatment, different effects and prognosis can be seen among squamous-cell carcinoma and adenocarcinoma. These differences indicate that there may be various methods of invasion and immunity between squamous-cell carcinoma and adenocarcinoma. Blood vessel invasion and tumor immune escape play very important roles in the progression and metastasis of cancer, and CD105 and integrins are novel therapeutic targets. We assessed the possible association of CD105 expression and integrins with TNM classification in patients with two types of NSCLC. A total of 72 patients with resected Non-Small Cell Lung Cancer (NSCLC) were reviewed retrospectively. Integrin β1, β2, β3, and α5β1 are assayed by immunofluorescence and integrin α5β1 using immunoblot. Intratumoral microvessel density was determined with an anti-CD34 mAb and an anti-CD105 mAb. Invasive ability was assayed with MMP2 and MMP9 using immunofluorescence. The expressions of all integrins, CD105, and CD34 are low in the normal lung tissue and highly expressed in the cancer niche compared to the adjacent tissues. CD105 is highly expressed in the adenocarcinoma niche compared to the squamous-cell carcinoma in NSCLC. The expressions of both MMP2 and MMP9 are low in the normal lung tissue and highly expressed in adjacent tissues. This study shows that blood vessel invasion appears to be an independent negative prognosticator in surgically managed types of NSCLC. However, adequately designed large prospective studies are warranted to confirm the present findings.     

The variation of the water deficit during the winter wheat growing season and its impact on crop yield in the North China Plain

The North China Plain (NCP) is one of the main agricultural areas in China. However, it is also widely known for its water shortages, especially during the winter wheat growing season. Recently, climate change has significantly affected the water environment for crop growth. Analyzing the changes in the water deficit, which is only affected by climate factor, will help to improve water management in the NCP. In this study, the Decision Support System for Agrotechnology Transfer (DSSAT) was used to investigate the variations in the water deficit during the winter wheat growing season from 1961 to 2010 in 12 selected stations in the NCP. To represent the changes in the water deficit without any artificial affection, the rainfed simulation was used. Over the past 50 years, the average temperature during the winter wheat growing season increased approximately 1.42 °C. The anthesis date moved forward approximately 7–10 days and to late April, which increased the water demand in April. Precipitation in March and May showed a positive trend, but there was a negative trend in April. The water deficit in late April and early May became more serious than before, with an increasing trend of more than 0.1 mm/year. In addition, because the heading stage, which is very important to crop yield of winter wheat, moved forward, the impact of water deficit in late April was more serious to crop yield.     

When Learners Surpass Their Models: Mathematical Modeling of Learning from an Inconsistent Source

It has been reported in the literature that both adults and children can, to a different degree, modify and regularize the often-inconsistent linguistic input they receive. We present a new algorithm to model and investigate the learning process of a learner mastering a set of (grammatical or lexical) forms from an inconsistent source. The algorithm is related to reinforcement learning and drift–diffusion models of decision making, and possesses several psychologically relevant properties such as fidelity, robustness, discounting, and computational simplicity. It demonstrates how a learner can successfully learn from or even surpass its imperfect source. We use the data collected by Singleton and Newport (Cognit Psychol 49(4):370–407, 2004) on the performance of a 7-year-boy Simon, who mastered the American Sign Language (ASL) by learning it from his parents, both of whom were imperfect speakers of ASL. We show that the algorithm possesses a frequency boosting property, whereby the frequency of the most common form of the source is increased by the learner. We also explain several key features of Simon’s ASL.     

Meta-analysis of the relationship between single nucleotide polymorphism rs72689236 of caspase-3 and Kawasaki disease

Kawasaki disease is a pediatric systemic vasculitis of unknown etiology, for which a genetic influence is suspected. But whether single nucleotide polymorphism (SNP) of caspase-3 rs72689236 is associated with Kawasaki disease is controversial. The aim of our study is to assess the association between the SNP of caspase-3 and risk for Kawasaki disease. We searched PubMed, MEDLINE, EMBASE, Springer, Elsevier Science Direct, Cochrane Library Google scholar, CNKI (China National Knowledge Infrastructure, in Chinese) and Wanfang database (in Chinese) to identify studies investigating the association between rs72689236 polymorphism and Kawasaki disease occurrence. There were five eligible studies, which included 4,241 (case group 1,560; control group 2,681) participants in this meta-analysis. Pooled odds ratios (ORs) and 95 % confidence intervals (95 % CIs) were calculated in a fixed-effects model (the Mantel–Haenszel method) or a random-effects model (the DerSimonian and Laird method) when appropriate. Significant associations were found under the overall ORs for A-allele comparison (A vs. G, pooled OR 1.33, 95 % CI 1.21–1.46), AA versus GG comparison (pooled OR 1.64, 95 % CI 1.35–2.00), GA versus GG comparison (pooled OR 1.42, 95 % CI 1.24–1.63), recessive model (AA vs. GG + GA, pooled OR 1.37, 95 % CI 1.15–1.64) and dominant model (AA + GA vs. GG, pooled OR 1.47, 95 % CI 1.29–1.67). This meta-analysis suggested that SNP rs72689236 of caspase-3 might be associated with susceptibility of Kawasaki disease and the allele A might increase the risk of Kawasaki disease in Asian samples such as Japanese and Chinese. In addition, individual studies with large sample size are needed to further evaluate the associations in various ethnic populations.     

Analyzing Kernel Matrices for the Identification of Differentially Expressed Genes

One of the most important applications of microarray data is the class prediction of biological samples. For this purpose, statistical tests have often been applied to identify the differentially expressed genes (DEGs), followed by the employment of the state-of-the-art learning machines including the Support Vector Machines (SVM) in particular. The SVM is a typical sample-based classifier whose performance comes down to how discriminant samples are. However, DEGs identified by statistical tests are not guaranteed to result in a training dataset composed of discriminant samples. To tackle this problem, a novel gene ranking method namely the Kernel Matrix Gene Selection (KMGS) is proposed. The rationale of the method, which roots in the fundamental ideas of the SVM algorithm, is described. The notion of ''''the separability of a sample'''' which is estimated by performing -like statistics on each column of the kernel matrix, is first introduced. The separability of a classification problem is then measured, from which the significance of a specific gene is deduced. Also described is a method of Kernel Matrix Sequential Forward Selection (KMSFS) which shares the KMGS method''s essential ideas but proceeds in a greedy manner. On three public microarray datasets, our proposed algorithms achieved noticeably competitive performance in terms of the B.632+ error rate.