Learning from Imbalanced Data

Methods for building QSARs

Structure-activity relationship (SAR) and quantitative structure-activity relationship (QSAR) models are increasingly used in toxicology, ecotoxicology, and pharmacology for predicting the activity of the molecules from their physicochemical properties and/or their structural characteristics. However, the design of such models has many traps for unwary practitioners. Consequently, the purpose of this chapter is to give a practical guide for the computation of SAR and QSAR models, point out problems that may be encountered, and suggest ways of solving them. Attempts are also made to see how these models can be validated and interpreted.

ECOH: an enzyme commission number predictor using mutual information and a support vector machine

MOTIVATION

The enzyme nomenclature system, commonly known as the enzyme commission (EC) number, plays a key role in classifying and predicting enzymatic reactions. However, numerous reactions have been described in various pathways that do not have an official EC number, and the reactions are not expected to have an EC number assigned because of a lack of articles published on enzyme assays. To predict the EC number of a non-classified enzymatic reaction, we focus on the structural similarity of its substrate and product to the substrate and product of reactions that have been classified.

RESULTS

We propose a new method to assign EC numbers using a maximum common substructure algorithm, mutual information and a support vector machine, termed the Enzyme COmmission numbers Handler (ECOH). A jack-knife test shows that the sensitivity, precision and accuracy of the method in predicting the first three digits of the official EC number (i.e. the EC sub-subclass) are 86.1%, 87.4% and 99.8%, respectively. We furthermore demonstrate that, by examining the ranking in the candidate lists of EC sub-subclasses generated by the algorithm, the method can successfully predict the classification of 85 enzymatic reactions that fall into multiple EC sub-subclasses. The better performance of the ECOH as compared with existing methods and its flexibility in predicting EC numbers make it useful for predicting enzyme function.

AVAILABILITY

ECOH is freely available via the Internet at http://www.bioinfo.sk.ritsumei.ac.jp/apps/ecoh/. This program only works on 32-bit Windows.

Advanced EEG analysis using threshold-free cluster-enhancement and non-parametric statistics

Advances in EEG signal analysis and its combination with other investigative techniques make appropriate statistical analysis of large EEG datasets a crucial issue. With an increasing number of available channels and samples, as well as more exploratory experimental designs, it has become necessary to develop a statistical process with a high level of statistical integrity, signal sensitivity which nonetheless produces results which are interpretable to the common user. Threshold-free cluster-enhancement has recently been proposed as a useful analysis tool for fMRI datasets. This approach essentially takes into account both a data point's statistical intensity and neighbourhood to transform the original signal into a more intuitive understanding of 'real' differences between groups or conditions. Here we adapt this approach to optimally deal with EEG datasets and use permutation-based statistics to build an efficient statistical analysis. Furthermore we compare the results with several other non-parametric and parametric approaches currently available using realistic simulated EEG signals. The proposed method is shown to be generally more sensitive to the variety of signal types common to EEG datasets without the need for any arbitrary adjusting of parameters. Moreover, a unique p-value is produced for each channel-sample pair such that specific questions can still be asked of the dataset while providing general information regarding the large-scale experimental effects.

SOMKE: kernel density estimation over data streams by sequences of self-organizing maps

In this paper, we propose a novel method SOMKE, for kernel density estimation (KDE) over data streams based on sequences of self-organizing map (SOM). In many stream data mining applications, the traditional KDE methods are infeasible because of the high computational cost, processing time, and memory requirement. To reduce the time and space complexity, we propose a SOM structure in this paper to obtain well-defined data clusters to estimate the underlying probability distributions of incoming data streams. The main idea of this paper is to build a series of SOMs over the data streams via two operations, that is, creating and merging the SOM sequences. The creation phase produces the SOM sequence entries for windows of the data, which obtains clustering information of the incoming data streams. The size of the SOM sequences can be further reduced by combining the consecutive entries in the sequence based on the measure of Kullback-Leibler divergence. Finally, the probability density functions over arbitrary time periods along the data streams can be estimated using such SOM sequences. We compare SOMKE with two other KDE methods for data streams, the M-kernel approach and the cluster kernel approach, in terms of accuracy and processing time for various stationary data streams. Furthermore, we also investigate the use of SOMKE over nonstationary (evolving) data streams, including a synthetic nonstationary data stream, a real-world financial data stream and a group of network traffic data streams. The simulation results illustrate the effectiveness and efficiency of the proposed approach.

A network-based feature selection approach to identify metabolic signatures in disease

The identification and interpretation of metabolic biomarkers is a challenging task. In this context, network-based approaches have become increasingly a key technology in systems biology allowing to capture complex interactions in biological systems. In this work, we introduce a novel network-based method to identify highly predictive biomarker candidates for disease. First, we infer two different types of networks: (i) correlation networks, and (ii) a new type of network called ratio networks. Based on these networks, we introduce scores to prioritize features using topological descriptors of the vertices. To evaluate our method we use an example dataset where quantitative targeted MS/MS analysis was applied to a total of 52 blood samples from 22 persons with obesity (BMI >30) and 30 healthy controls. Using our network-based feature selection approach we identified highly discriminating metabolites for obesity (F-score >0.85, accuracy >85%), some of which could be verified by the literature.

SSC: a classifier combination method based on signal strength

We propose a new classifier combination method, the signal strength-based combining (SSC) approach, to combine the outputs of multiple classifiers to support the decision-making process in classification tasks. As ensemble learning methods have attracted growing attention from both academia and industry recently, it is critical to understand the fundamental issues of the combining rule. Motivated by the signal strength concept, our proposed SSC algorithm can effectively integrate the individual vote from different classifiers in an ensemble learning system. Comparative studies of our method with nine major existing combining rules, namely, geometric average rule, arithmetic average rule, median value rule, majority voting rule, Borda count, max and min rule, weighted average, and weighted majority voting rules, is presented. Furthermore, we also discuss the relationship of the proposed method with respect to margin-based classifiers, including the boosting method (AdaBoost.M1 and AdaBoost.M2) and support vector machines by margin analysis. Detailed analyses of margin distribution graphs are presented to discuss the characteristics of the proposed method. Simulation results for various real-world datasets illustrate the effectiveness of the proposed method.

Performance evaluation of multilayer perceptrons for discriminating and quantifying multiple kinds of odors with an electronic nose

This paper studies several types and arrangements of perceptron modules to discriminate and quantify multiple odors with an electronic nose. We evaluate the following types of multilayer perceptron. (A) A single multi-output (SMO) perceptron both for discrimination and for quantification. (B) An SMO perceptron for discrimination followed by multiple multi-output (MMO) perceptrons for quantification. (C) An SMO perceptron for discrimination followed by multiple single-output (MSO) perceptrons for quantification. (D) MSO perceptrons for discrimination followed by MSO perceptrons for quantification, called the MSO-MSO perceptron model, under the following conditions: (D1) using a simple one-against-all (OAA) decomposition method; (D2) adopting a simple OAA decomposition method and virtual balance step; and (D3) employing a local OAA decomposition method, virtual balance step and local generalization strategy all together. The experimental results for 12 kinds of volatile organic compounds at 85 concentration levels in the training set and 155 concentration levels in the test set show that the MSO-MSO perceptron model with the D3 learning procedure is the most effective of those tested for discrimination and quantification of many kinds of odors.

Predicting protein-ATP binding sites from primary sequence through fusing bi-profile sampling of multi-view features

Background

Adenosine-5′-triphosphate (ATP) is one of multifunctional nucleotides and plays an important role in cell biology as a coenzyme interacting with proteins. Revealing the binding sites between protein and ATP is significantly important to understand the functionality of the proteins and the mechanisms of protein-ATP complex.

Results

In this paper, we propose a novel framework for predicting the proteins’ functional residues, through which they can bind with ATP molecules. The new prediction protocol is achieved by combination of sequence evolutional information and bi-profile sampling of multi-view sequential features and the sequence derived structural features. The hypothesis for this strategy is single-view feature can only represent partial target’s knowledge and multiple sources of descriptors can be complementary.

Conclusions

Prediction performances evaluated by both 5-fold and leave-one-out jackknife cross-validation tests on two benchmark datasets consisting of 168 and 227 non-homologous ATP binding proteins respectively demonstrate the efficacy of the proposed protocol. Our experimental results also reveal that the residue structural characteristics of real protein-ATP binding sites are significant different from those normal ones, for example the binding residues do not show high solvent accessibility propensities, and the bindings prefer to occur at the conjoint points between different secondary structure segments. Furthermore, results also show that performance is affected by the imbalanced training datasets by testing multiple ratios between positive and negative samples in the experiments. Increasing the dataset scale is also demonstrated useful for improving the prediction performances.

Improving the prediction of sub-cellular locations of proteins with a particle swarm optimization-based boosting strategy

Learning from imbalanced data sets presents an important challenge to the machine learning community. Traditional classification methods, seeking to minimize the overall error rate of the whole training set, do not perform well on imbalanced data since they assume a relatively balanced class distribution and put too much strength on the majority class. This is a common scenario when predicting sub-cellular locations of proteins since proteins belonging to certain specific locations are naturally more abundant or have been more extensively studied. In this work, a new method to learn from imbalanced data, called SwarmBoost, is proposed in order to reduce overlapping and noise of imbalanced datasets and improve prediction performances. The method combines oversampling, subsampling based on particle swarm optimization and ensemble methods. Our results show that SwarmBoost equals and in several cases outperforms other common boosting algorithms like DataBoost-Im and AdaBoost, constituting a useful tool for improving sub-cellular location predictions.

Identifying Rare and Subtle Behaviors: A Weakly Supervised Joint Topic Model

One of the most interesting and desired capabilities for automated video behavior analysis is the identification of rarely occurring and subtle behaviors. This is of practical value because dangerous or illegal activities often have few or possibly only one prior example to learn from and are often subtle. Rare and subtle behavior learning is challenging for two reasons: (1) Contemporary modeling approaches require more data and supervision than may be available and (2) the most interesting and potentially critical rare behaviors are often visually subtle-occurring among more obvious typical behaviors or being defined by only small spatio-temporal deviations from typical behaviors. In this paper, we introduce a novel weakly supervised joint topic model which addresses these issues. Specifically, we introduce a multiclass topic model with partially shared latent structure and associated learning and inference algorithms. These contributions will permit modeling of behaviors from as few as one example, even without localization by the user and when occurring in clutter, and subsequent classification and localization of such behaviors online and in real time. We extensively validate our approach on two standard public-space data sets, where it clearly outperforms a batch of contemporary alternatives.

Improving ECG classification accuracy using an ensemble of neural network modules

This paper illustrates the use of a combined neural network model based on Stacked Generalization method for classification of electrocardiogram (ECG) beats. In conventional Stacked Generalization method, the combiner learns to map the base classifiers' outputs to the target data. We claim adding the input pattern to the base classifiers' outputs helps the combiner to obtain knowledge about the input space and as the result, performs better on the same task. Experimental results support our claim that the additional knowledge according to the input space, improves the performance of the proposed method which is called Modified Stacked Generalization. In particular, for classification of 14966 ECG beats that were not previously seen during training phase, the Modified Stacked Generalization method reduced the error rate for 12.41% in comparison with the best of ten popular classifier fusion methods including Max, Min, Average, Product, Majority Voting, Borda Count, Decision Templates, Weighted Averaging based on Particle Swarm Optimization and Stacked Generalization.

Automatic defect detection for TFT-LCD array process using quasiconformal kernel support vector data description

Defect detection has been considered an efficient way to increase the yield rate of panels in thin film transistor liquid crystal display (TFT-LCD) manufacturing. In this study we focus on the array process since it is the first and key process in TFT-LCD manufacturing. Various defects occur in the array process, and some of them could cause great damage to the LCD panels. Thus, how to design a method that can robustly detect defects from the images captured from the surface of LCD panels has become crucial. Previously, support vector data description (SVDD) has been successfully applied to LCD defect detection. However, its generalization performance is limited. In this paper, we propose a novel one-class machine learning method, called quasiconformal kernel SVDD (QK-SVDD) to address this issue. The QK-SVDD can significantly improve generalization performance of the traditional SVDD by introducing the quasiconformal transformation into a predefined kernel. Experimental results, carried out on real LCD images provided by an LCD manufacturer in Taiwan, indicate that the proposed QK-SVDD not only obtains a high defect detection rate of 96%, but also greatly improves generalization performance of SVDD. The improvement has shown to be over 30%. In addition, results also show that the QK-SVDD defect detector is able to accomplish the task of defect detection on an LCD image within 60 ms.

An evaluation protocol for subtype-specific breast cancer event prediction

In recent years increasing evidence appeared that breast cancer may not constitute a single disease at the molecular level, but comprises a heterogeneous set of subtypes. This suggests that instead of building a single monolithic predictor, better predictors might be constructed that solely target samples of a designated subtype, which are believed to represent more homogeneous sets of samples. An unavoidable drawback of developing subtype-specific predictors, however, is that a stratification by subtype drastically reduces the number of samples available for their construction. As numerous studies have indicated sample size to be an important factor in predictor construction, it is therefore questionable whether the potential benefit of subtyping can outweigh the drawback of a severe loss in sample size. Factors like unequal class distributions and differences in the number of samples per subtype, further complicate comparisons. We present a novel experimental protocol that facilitates a comprehensive comparison between subtype-specific predictors and predictors that do not take subtype information into account. Emphasis lies on careful control of sample size as well as class and subtype distributions. The methodology is applied to a large breast cancer compendium involving over 1500 arrays, using a state-of-the-art subtyping scheme. We show that the resulting subtype-specific predictors outperform those that do not take subtype information into account, especially when taking sample size considerations into account.

Prediction of leucine-rich nuclear export signal containing proteins with NESsential

The classical nuclear export signal (NES), also known as the leucine-rich NES, is a protein localization signal often involved in important processes such as signal transduction and cell cycle regulation. Although 15 years has passed since its discovery, limited structural information and high sequence diversity have hampered understanding of the NES. Several consensus sequences have been proposed to describe it, but they suffer from poor predictive power. On the other hand, the NetNES server provides the only computational method currently available. Although these two methods have been widely used to attempt to find the correct NES position within potential NES-containing proteins, their performance has not yet been evaluated on the basic task of identifying NES-containing proteins. We propose a new predictor, NESsential, which uses sequence derived meta-features, such as predicted disorder and solvent accessibility, in addition to primary sequence. We demonstrate that it can identify promising NES-containing candidate proteins (albeit at low coverage), but other methods cannot. We also quantitatively demonstrate that predicted disorder is a useful feature for prediction and investigate the different features of (predicted) ordered versus disordered NES's. Finally, we list 70 recently discovered NES-containing proteins, doubling the number available to the community.

Novel computational identification of highly selective biomarkers of pollutant exposure

The use of in vivo biosensors to acquire environmental pollution data is an emerging and promising paradigm. One major challenge is the identification of highly specific biomarkers that selectively report exposure to a target pollutant, while remaining quiescent under a diverse set of other, often unknown, environmental conditions. This study hypothesized that a microarray data mining approach can identify highly specific biomarkers, and, that the robustness property can generalize to unforeseen environmental conditions. Starting with Arabidopsis thaliana microarray data measuring responses to a variety of treatments, the study used the top scoring pair (TSP) algorithm to identify mRNA transcripts that respond uniquely to phenanthrene, a model polycyclic aromatic hydrocarbon. Subsequent in silico analysis with a larger set of microarray data indicated that the biomarkers remained robust under new conditions. Finally, in vivo experiments were performed with unforeseen conditions that mimic phenanthrene stress, and the biomarkers were assayed using qRT-PCR. In these experiments, the biomarkers always responded positively to phenanthrene, and never responded to the unforeseen conditions, thereby supporting the hypotheses. This data mining approach requires only microarray or next-generation RNA-seq data, and, in principle, can be applied to arbitrary biomonitoring organisms and chemical exposures.