An Unsupervised Text Mining Method for Relation Extraction from Biomedical Literature

The wealth of interaction information provided in biomedical articles motivated the implementation of text mining approaches to automatically extract biomedical relations. This paper presents an unsupervised method based on pattern clustering and sentence parsing to deal with biomedical relation extraction. Pattern clustering algorithm is based on Polynomial Kernel method, which identifies interaction words from unlabeled data; these interaction words are then used in relation extraction between entity pairs. Dependency parsing and phrase structure parsing are combined for relation extraction. Based on the semi-supervised KNN algorithm, we extend the proposed unsupervised approach to a semi-supervised approach by combining pattern clustering, dependency parsing and phrase structure parsing rules. We evaluated the approaches on two different tasks: (1) Protein–protein interactions extraction, and (2) Gene–suicide association extraction. The evaluation of task (1) on the benchmark dataset (AImed corpus) showed that our proposed unsupervised approach outperformed three supervised methods. The three supervised methods are rule based, SVM based, and Kernel based separately. The proposed semi-supervised approach is superior to the existing semi-supervised methods. The evaluation on gene–suicide association extraction on a smaller dataset from Genetic Association Database and a larger dataset from publicly available PubMed showed that the proposed unsupervised and semi-supervised methods achieved much higher F-scores than co-occurrence based method.     

CTANet: Confidence-Based Threshold Adaption Network for Semi-Supervised Segmentation of Uterine Regions from MR Images for HIFU Treatment

ObjectivesThe accurate preoperative segmentation of the uterus and uterine fibroids from magnetic resonance images (MRI) is an essential step for diagnosis and real-time ultrasound guidance during high-intensity focused ultrasound (HIFU) surgery. Conventional supervised methods are effective techniques for image segmentation. Recently, semi-supervised segmentation approaches have been reported in the literature. One popular technique for semi-supervised methods is to use pseudo-labels to artificially annotate unlabeled data. However, many existing pseudo-label generations rely on a fixed threshold used to generate a confidence map, regardless of the proportion of unlabeled and labeled data.Materials and MethodsTo address this issue, we propose a novel semi-supervised framework called Confidence-based Threshold Adaptation Network (CTANet) to improve the quality of pseudo-labels. Specifically, we propose an online pseudo-labels method to automatically adjust the threshold, producing high-confident unlabeled annotations and boosting segmentation accuracy. To further improve the network's generalization to fit the diversity of different patients, we design a novel mixup strategy by regularizing the network on each layer in the decoder part and introducing a consistency regularization loss between the outputs of two sub-networks in CTANet.ResultsWe compare our method with several state-of-the-art semi-supervised segmentation methods on the same uterine fibroids dataset containing 297 patients. The performance is evaluated by the Dice similarity coefficient, the precision, and the recall. The results show that our method outperforms other semi-supervised learning methods. Moreover, for the same training set, our method approaches the segmentation performance of a fully supervised U-Net (100% annotated data) but using 4 times less annotated data (25% annotated data, 75% unannotated data).ConclusionExperimental results are provided to illustrate the effectiveness of the proposed semi-supervised approach. The proposed method can contribute to multi-class segmentation of uterine regions from MRI for HIFU treatment.     

The S(2)-Ensemble Fusion Algorithm

This paper presents a novel model for performing classification and visualization of high-dimensional data by means of combining two enhancing techniques. The first is a semi-supervised learning, an extension of the supervised learning used to incorporate unlabeled information to the learning process. The second is an ensemble learning to replicate the analysis performed, followed by a fusion mechanism that yields as a combined result of previously performed analysis in order to improve the result of a single model. The proposed learning schema, termed S(2)-Ensemble, is applied to several unsupervised learning algorithms within the family of topology maps, such as the Self-Organizing Maps and the Neural Gas. This study also includes a thorough research of the characteristics of these novel schemes, by means quality measures, which allow a complete analysis of the resultant classifiers from the viewpoint of various perspectives over the different ways that these classifiers are used. The study conducts empirical evaluations and comparisons on various real-world datasets from the UCI repository, which exhibit different characteristics, so to enable an extensive selection of situations where the presented new algorithms can be applied.     

Laplacian Linear Discriminant Analysis Approach to Unsupervised Feature Selection

Until recently, numerous feature selection techniques have been proposed and found wide applications in genomics and proteomics. For instance, feature/gene selection has proven to be useful for biomarker discovery from microarray and mass spectrometry data. While supervised feature selection has been explored extensively, there are only a few unsupervised methods that can be applied to exploratory data analysis. In this paper, we address the problem of unsupervised feature selection. First, we extend Laplacian linear discriminant analysis (LLDA) to unsupervised cases. Second, we propose a novel algorithm for computing LLDA, which is efficient in the case of high dimensionality and small sample size as in microarray data. Finally, an unsupervised feature selection method, called LLDA-based Recursive Feature Elimination (LLDA-RFE), is proposed. We apply LLDA-RFE to several public data sets of cancer microarrays and compare its performance with those of Laplacian score and SVD-entropy, two state-of-the-art unsupervised methods, and with that of Fisher score, a supervised filter method. Our results demonstrate that LLDA-RFE outperforms Laplacian score and shows favorable performance against SVD-entropy. It performs even better than Fisher score for some of the data sets, despite the fact that LLDA-RFE is fully unsupervised.     

Mining Proteins with Non-Experimental Annotations Based on an Active Sample Selection Strategy for Predicting Protein Subcellular Localization

Subcellular localization of a protein is important to understand proteins’ functions and interactions. There are many techniques based on computational methods to predict protein subcellular locations, but it has been shown that many prediction tasks have a training data shortage problem. This paper introduces a new method to mine proteins with non-experimental annotations, which are labeled by non-experimental evidences of protein databases to overcome the training data shortage problem. A novel active sample selection strategy is designed, taking advantage of active learning technology, to actively find useful samples from the entire data pool of candidate proteins with non-experimental annotations. This approach can adequately estimate the “value” of each sample, automatically select the most valuable samples and add them into the original training set, to help to retrain the classifiers. Numerical experiments with for four popular multi-label classifiers on three benchmark datasets show that the proposed method can effectively select the valuable samples to supplement the original training set and significantly improve the performances of predicting classifiers.     

Hierarchical Gene Selection and Genetic Fuzzy System for Cancer Microarray Data Classification

This paper introduces a novel approach to gene selection based on a substantial modification of analytic hierarchy process (AHP). The modified AHP systematically integrates outcomes of individual filter methods to select the most informative genes for microarray classification. Five individual ranking methods including t-test, entropy, receiver operating characteristic (ROC) curve, Wilcoxon and signal to noise ratio are employed to rank genes. These ranked genes are then considered as inputs for the modified AHP. Additionally, a method that uses fuzzy standard additive model (FSAM) for cancer classification based on genes selected by AHP is also proposed in this paper. Traditional FSAM learning is a hybrid process comprising unsupervised structure learning and supervised parameter tuning. Genetic algorithm (GA) is incorporated in-between unsupervised and supervised training to optimize the number of fuzzy rules. The integration of GA enables FSAM to deal with the high-dimensional-low-sample nature of microarray data and thus enhance the efficiency of the classification. Experiments are carried out on numerous microarray datasets. Results demonstrate the performance dominance of the AHP-based gene selection against the single ranking methods. Furthermore, the combination of AHP-FSAM shows a great accuracy in microarray data classification compared to various competing classifiers. The proposed approach therefore is useful for medical practitioners and clinicians as a decision support system that can be implemented in the real medical practice.     

AVC: Selecting discriminative features on basis of AUC by maximizing variable complementarity

Background

The Receiver Operator Characteristic (ROC) curve is well-known in evaluating classification performance in biomedical field. Owing to its superiority in dealing with imbalanced and cost-sensitive data, the ROC curve has been exploited as a popular metric to evaluate and find out disease-related genes (features). The existing ROC-based feature selection approaches are simple and effective in evaluating individual features. However, these approaches may fail to find real target feature subset due to their lack of effective means to reduce the redundancy between features, which is essential in machine learning.

Results

In this paper, we propose to assess feature complementarity by a trick of measuring the distances between the misclassified instances and their nearest misses on the dimensions of pairwise features. If a misclassified instance and its nearest miss on one feature dimension are far apart on another feature dimension, the two features are regarded as complementary to each other. Subsequently, we propose a novel filter feature selection approach on the basis of the ROC analysis. The new approach employs an efficient heuristic search strategy to select optimal features with highest complementarities. The experimental results on a broad range of microarray data sets validate that the classifiers built on the feature subset selected by our approach can get the minimal balanced error rate with a small amount of significant features.

Conclusions

Compared with other ROC-based feature selection approaches, our new approach can select fewer features and effectively improve the classification performance.

     

Semi-supervised analysis of gene expression profiles for lineage-specific development in the Caenorhabditis elegans embryo

MOTIVATION: Gene expression profiling is a powerful approach to identify genes that may be involved in a specific biological process on a global scale. For example, gene expression profiling of mutant animals that lack or contain an excess of certain cell types is a common way to identify genes that are important for the development and maintenance of given cell types. However, it is difficult for traditional computational methods, including unsupervised and supervised learning methods, to detect relevant genes from a large collection of expression profiles with high sensitivity and specificity. Unsupervised methods group similar gene expressions together while ignoring important prior biological knowledge. Supervised methods utilize training data from prior biological knowledge to classify gene expression. However, for many biological problems, little prior knowledge is available, which limits the prediction performance of most supervised methods. RESULTS: We present a Bayesian semi-supervised learning method, called BGEN, that improves upon supervised and unsupervised methods by both capturing relevant expression profiles and using prior biological knowledge from literature and experimental validation. Unlike currently available semi-supervised learning methods, this new method trains a kernel classifier based on labeled and unlabeled gene expression examples. The semi-supervised trained classifier can then be used to efficiently classify the remaining genes in the dataset. Moreover, we model the confidence of microarray probes and probabilistically combine multiple probe predictions into gene predictions. We apply BGEN to identify genes involved in the development of a specific cell lineage in the C. elegans embryo, and to further identify the tissues in which these genes are enriched. Compared to K-means clustering and SVM classification, BGEN achieves higher sensitivity and specificity. We confirm certain predictions by biological experiments. AVAILABILITY: The results are available at http://www.csail.mit.edu/~alanqi/projects/BGEN.html.     

R package for animal behavior classification from accelerometer data—rabc

Increasingly, animal behavior studies are enhanced through the use of accelerometry. To allow translation of raw accelerometer data to animal behaviors requires the development of classifiers. Here, we present the “rabc” (r for animal behavior classification) package to assist researchers with the interactive development of such animal behavior classifiers in a supervised classification approach. The package uses datasets consisting of accelerometer data with their corresponding animal behaviors (e.g., for triaxial accelerometer data along the x, y and z axes arranged as “x, y, z, x, y, z,…, behavior”). Using an example dataset collected on white stork (Ciconia ciconia), we illustrate the workflow of this package, including accelerometer data visualization, feature calculation, feature selection, feature visualization, extreme gradient boost model training, validation, and, finally, a demonstration of the behavior classification results.     

Benign-malignant pulmonary nodule classification in low-dose CT with convolutional features

PurposeLow-Dose Computed Tomography (LDCT) is the most common imaging modality for lung cancer diagnosis. The presence of nodules in the scans does not necessarily portend lung cancer, as there is an intricate relationship between nodule characteristics and lung cancer. Therefore, benign-malignant pulmonary nodule classification at early detection is a crucial step to improve diagnosis and prolong patient survival. The aim of this study is to propose a method for predicting nodule malignancy based on deep abstract features.MethodsTo efficiently capture both intra-nodule heterogeneities and contextual information of the pulmonary nodules, a dual pathway model was developed to integrate the intra-nodule characteristics with contextual attributes. The proposed approach was implemented with both supervised and unsupervised learning schemes. A random forest model was added as a second component on top of the networks to generate the classification results. The discrimination power of the model was evaluated by calculating the Area Under the Receiver Operating Characteristic Curve (AUROC) metric.ResultsExperiments on 1297 manually segmented nodules show that the integration of context and target supervised deep features have a great potential for accurate prediction, resulting in a discrimination power of 0.936 in terms of AUROC, which outperformed the classification performance of the Kaggle 2017 challenge winner.ConclusionEmpirical results demonstrate that integrating nodule target and context images into a unified network improves the discrimination power, outperforming the conventional single pathway convolutional neural networks.     

A blocking strategy to improve gene selection for classification of gene expression data

Because of high dimensionality, machine learning algorithms typically rely on feature selection techniques in order to perform effective classification in microarray gene expression data sets. However, the large number of features compared to the number of samples makes the task of feature selection computationally hard and prone to errors. This paper interprets feature selection as a task of stochastic optimization, where the goal is to select among an exponential number of alternative gene subsets the one expected to return the highest generalization in classification. Blocking is an experimental design strategy which produces similar experimental conditions to compare alternative stochastic configurations in order to be confident that observed differences in accuracy are due to actual differences rather than to fluctuations and noise effects. We propose an original blocking strategy for improving feature selection which aggregates in a paired way the validation outcomes of several learning algorithms to assess a gene subset and compare it to others. This is a novelty with respect to conventional wrappers, which commonly adopt a sole learning algorithm to evaluate the relevance of a given set of variables. The rationale of the approach is that, by increasing the amount of experimental conditions under which we validate a feature subset, we can lessen the problems related to the scarcity of samples and consequently come up with a better selection. The paper shows that the blocking strategy significantly improves the performance of a conventional forward selection for a set of 16 publicly available cancer expression data sets. The experiments involve six different classifiers and show that improvements take place independent of the classification algorithm used after the selection step. Two further validations based on available biological annotation support the claim that blocking strategies in feature selection may improve the accuracy and the quality of the solution. The first validation is based on retrieving PubMEd abstracts associated to the selected genes and matching them to regular expressions describing the biological phenomenon underlying the expression data sets. The biological validation that follows is based on the use of the Bioconductor package GoStats in order to perform Gene Ontology statistical analysis.     

Inferring mouse gene functions from genomic-scale data using a combined functional network/classification strategy

The complete set of mouse genes, as with the set of human genes, is still largely uncharacterized, with many pieces of experimental evidence accumulating regarding the activities and expression of the genes, but the majority of genes as yet still of unknown function. Within the context of the MouseFunc competition, we developed and applied two distinct large-scale data mining approaches to infer the functions (Gene Ontology annotations) of mouse genes from experimental observations from available functional genomics, proteomics, comparative genomics, and phenotypic data. The two strategies - the first using classifiers to map features to annotations, the second propagating annotations from characterized genes to uncharacterized genes along edges in a network constructed from the features - offer alternative and possibly complementary approaches to providing functional annotations. Here, we re-implement and evaluate these approaches and their combination for their ability to predict the proper functional annotations of genes in the MouseFunc data set. We show that, when controlling for the same set of input features, the network approach generally outperformed a naive Bayesian classifier approach, while their combination offers some improvement over either independently. We make our observations of predictive performance on the MouseFunc competition hold-out set, as well as on a ten-fold cross-validation of the MouseFunc data. Across all 1,339 annotated genes in the MouseFunc test set, the median predictive power was quite strong (median area under a receiver operating characteristic plot of 0.865 and average precision of 0.195), indicating that a mining-based strategy with existing data is a promising path towards discovering mammalian gene functions. As one product of this work, a high-confidence subset of the functional mouse gene network was produced - spanning >70% of mouse genes with >1.6 million associations - that is predictive of mouse (and therefore often human) gene function and functional associations. The network should be generally useful for mammalian gene functional analyses, such as for predicting interactions, inferring functional connections between genes and pathways, and prioritizing candidate genes. The network and all predictions are available on the worldwide web.     

Heterogeneous Ensemble Combination Search Using Genetic Algorithm for Class Imbalanced Data Classification

Classification of datasets with imbalanced sample distributions has always been a challenge. In general, a popular approach for enhancing classification performance is the construction of an ensemble of classifiers. However, the performance of an ensemble is dependent on the choice of constituent base classifiers. Therefore, we propose a genetic algorithm-based search method for finding the optimum combination from a pool of base classifiers to form a heterogeneous ensemble. The algorithm, called GA-EoC, utilises 10 fold-cross validation on training data for evaluating the quality of each candidate ensembles. In order to combine the base classifiers decision into ensemble’s output, we used the simple and widely used majority voting approach. The proposed algorithm, along with the random sub-sampling approach to balance the class distribution, has been used for classifying class-imbalanced datasets. Additionally, if a feature set was not available, we used the (α, β) − k Feature Set method to select a better subset of features for classification. We have tested GA-EoC with three benchmarking datasets from the UCI-Machine Learning repository, one Alzheimer’s disease dataset and a subset of the PubFig database of Columbia University. In general, the performance of the proposed method on the chosen datasets is robust and better than that of the constituent base classifiers and many other well-known ensembles. Based on our empirical study we claim that a genetic algorithm is a superior and reliable approach to heterogeneous ensemble construction and we expect that the proposed GA-EoC would perform consistently in other cases.     

Identification of novel RNA design candidates by clustering the extended RNA-As-Graphs library

BackgroundWe re-evaluate our RNA-As-Graphs clustering approach, using our expanded graph library and new RNA structures, to identify potential RNA-like topologies for design. Our coarse-grained approach represents RNA secondary structures as tree and dual graphs, with vertices and edges corresponding to RNA helices and loops. The graph theoretical framework facilitates graph enumeration, partitioning, and clustering approaches to study RNA structure and its applications.MethodsClustering graph topologies based on features derived from graph Laplacian matrices and known RNA structures allows us to classify topologies into ‘existing’ or hypothetical, and the latter into, ‘RNA-like’ or ‘non RNA-like’ topologies. Here we update our list of existing tree graph topologies and RAG-3D database of atomic fragments to include newly determined RNA structures. We then use linear and quadratic regression, optionally with dimensionality reduction, to derive graph features and apply several clustering algorithms on our tree-graph library and recently expanded dual-graph library to classify them into the three groups.ResultsThe unsupervised PAM and K-means clustering approaches correctly classify 72–77% of all existing graph topologies and 75–82% of newly added ones as RNA-like. For supervised k-NN clustering, the cross-validation accuracy ranges from 57 to 81%.ConclusionsUsing linear regression with unsupervised clustering, or quadratic regression with supervised clustering, provides better accuracies than supervised/linear clustering. All accuracies are better than random, especially for newly added existing topologies, thus lending credibility to our approach.General significanceOur updated RAG-3D database and motif classification by clustering present new RNA substructures and RNA-like motifs as novel design candidates.     

Supervised and Unsupervised Self-Testing for HIV in High- and Low-Risk Populations: A Systematic Review

Background

Stigma, discrimination, lack of privacy, and long waiting times partly explain why six out of ten individuals living with HIV do not access facility-based testing. By circumventing these barriers, self-testing offers potential for more people to know their sero-status. Recent approval of an in-home HIV self test in the US has sparked self-testing initiatives, yet data on acceptability, feasibility, and linkages to care are limited. We systematically reviewed evidence on supervised (self-testing and counselling aided by a health care professional) and unsupervised (performed by self-tester with access to phone/internet counselling) self-testing strategies.

Methods and Findings

Seven databases (Medline [via PubMed], Biosis, PsycINFO, Cinahl, African Medicus, LILACS, and EMBASE) and conference abstracts of six major HIV/sexually transmitted infections conferences were searched from 1st January 2000–30th October 2012. 1,221 citations were identified and 21 studies included for review. Seven studies evaluated an unsupervised strategy and 14 evaluated a supervised strategy. For both strategies, data on acceptability (range: 74%–96%), preference (range: 61%–91%), and partner self-testing (range: 80%–97%) were high. A high specificity (range: 99.8%–100%) was observed for both strategies, while a lower sensitivity was reported in the unsupervised (range: 92.9%–100%; one study) versus supervised (range: 97.4%–97.9%; three studies) strategy. Regarding feasibility of linkage to counselling and care, 96% (n = 102/106) of individuals testing positive for HIV stated they would seek post-test counselling (unsupervised strategy, one study). No extreme adverse events were noted. The majority of data (n = 11,019/12,402 individuals, 89%) were from high-income settings and 71% (n = 15/21) of studies were cross-sectional in design, thus limiting our analysis.

Conclusions

Both supervised and unsupervised testing strategies were highly acceptable, preferred, and more likely to result in partner self-testing. However, no studies evaluated post-test linkage with counselling and treatment outcomes and reporting quality was poor. Thus, controlled trials of high quality from diverse settings are warranted to confirm and extend these findings. Please see later in the article for the Editors'' Summary     

Machine learning-based risk factor analysis and prevalence prediction of intestinal parasitic infections using epidemiological survey data

BackgroundPrevious epidemiological studies have examined the prevalence and risk factors for a variety of parasitic illnesses, including protozoan and soil-transmitted helminth (STH, e.g., hookworms and roundworms) infections. Despite advancements in machine learning for data analysis, the majority of these studies use traditional logistic regression to identify significant risk factors.MethodsIn this study, we used data from a survey of 54 risk factors for intestinal parasitosis in 954 Ethiopian school children. We investigated whether machine learning approaches can supplement traditional logistic regression in identifying intestinal parasite infection risk factors. We used feature selection methods such as InfoGain (IG), ReliefF (ReF), Joint Mutual Information (JMI), and Minimum Redundancy Maximum Relevance (MRMR). Additionally, we predicted children’s parasitic infection status using classifiers such as Logistic Regression (LR), Support Vector Machines (SVM), Random Forests (RF) and XGBoost (XGB), and compared their accuracy and area under the receiver operating characteristic curve (AUROC) scores. For optimal model training, we performed tenfold cross-validation and tuned the classifier hyperparameters. We balanced our dataset using the Synthetic Minority Oversampling (SMOTE) method. Additionally, we used association rule learning to establish a link between risk factors and parasitic infections.Key findingsOur study demonstrated that machine learning could be used in conjunction with logistic regression. Using machine learning, we developed models that accurately predicted four parasitic infections: any parasitic infection at 79.9% accuracy, helminth infection at 84.9%, any STH infection at 95.9%, and protozoan infection at 94.2%. The Random Forests (RF) and Support Vector Machines (SVM) classifiers achieved the highest accuracy when top 20 risk factors were considered using Joint Mutual Information (JMI) or all features were used. The best predictors of infection were socioeconomic, demographic, and hematological characteristics.ConclusionsWe demonstrated that feature selection and association rule learning are useful strategies for detecting risk factors for parasite infection. Additionally, we showed that advanced classifiers might be utilized to predict children’s parasitic infection status. When combined with standard logistic regression models, machine learning techniques can identify novel risk factors and predict infection risk.     

Gene expression data analysis with a dynamically extended self-organized map that exploits class information

MOTIVATION: Currently the most popular approach to analyze genome-wide expression data is clustering. One of the major drawbacks of most of the existing clustering methods is that the number of clusters has to be specified a priori. Furthermore, by using pure unsupervised algorithms prior biological knowledge is totally ignored Moreover, most current tools lack an effective framework for tight integration of unsupervised and supervised learning for the analysis of high-dimensional expression data and only very few multi-class supervised approaches are designed with the provision for effectively utilizing multiple functional class labeling. RESULTS: The paper adapts a novel Self-Organizing map called supervised Network Self-Organized Map (sNet-SOM) to the peculiarities of multi-labeled gene expression data. The sNet-SOM determines adaptively the number of clusters with a dynamic extension process. This process is driven by an inhomogeneous measure that tries to balance unsupervised, supervised and model complexity criteria. Nodes within a rectangular grid are grown at the boundary nodes, weights rippled from the internal nodes towards the outer nodes of the grid, and whole columns inserted within the map The appropriate level of expansion is determined automatically. Multiple sNet-SOM models are constructed dynamically each for a different unsupervised/supervised balance and model selection criteria are used to select the one optimum one. The results indicate that sNet-SOM yields competitive performance to other recently proposed approaches for supervised classification at a significantly reduced computational cost and it provides extensive exploratory analysis potentiality within the analysis framework. Furthermore, it explores simple design decisions that are easier to comprehend and computationally efficient.     

On the Relevance of Sophisticated Structural Annotations for Disulfide Connectivity Pattern Prediction

Disulfide bridges strongly constrain the native structure of many proteins and predicting their formation is therefore a key sub-problem of protein structure and function inference. Most recently proposed approaches for this prediction problem adopt the following pipeline: first they enrich the primary sequence with structural annotations, second they apply a binary classifier to each candidate pair of cysteines to predict disulfide bonding probabilities and finally, they use a maximum weight graph matching algorithm to derive the predicted disulfide connectivity pattern of a protein. In this paper, we adopt this three step pipeline and propose an extensive study of the relevance of various structural annotations and feature encodings. In particular, we consider five kinds of structural annotations, among which three are novel in the context of disulfide bridge prediction. So as to be usable by machine learning algorithms, these annotations must be encoded into features. For this purpose, we propose four different feature encodings based on local windows and on different kinds of histograms. The combination of structural annotations with these possible encodings leads to a large number of possible feature functions. In order to identify a minimal subset of relevant feature functions among those, we propose an efficient and interpretable feature function selection scheme, designed so as to avoid any form of overfitting. We apply this scheme on top of three supervised learning algorithms: k-nearest neighbors, support vector machines and extremely randomized trees. Our results indicate that the use of only the PSSM (position-specific scoring matrix) together with the CSP (cysteine separation profile) are sufficient to construct a high performance disulfide pattern predictor and that extremely randomized trees reach a disulfide pattern prediction accuracy of on the benchmark dataset SPX, which corresponds to improvement over the state of the art. A web-application is available at http://m24.giga.ulg.ac.be:81/x3CysBridges.