Global feature selection from microarray data using Lagrange multipliers

Identifying discriminative features for diagnosis of Kashin-Beck disease among adolescents

INTRODUCTION

Diagnosing Kashin-Beck disease (KBD) involves damages to multiple joints and carries variable clinical symptoms, posing great challenge to the diagnosis of KBD for clinical practitioners. However, it is still unclear which clinical features of KBD are more informative for the diagnosis of Kashin-Beck disease among adolescent.

METHODS

We first manually extracted 26 possible features including clinical manifestations, and pathological changes of X-ray images from 400 KBD and 400 non-KBD adolescents. With such features, we performed four classification methods, i.e., random forest algorithms (RFA), artificial neural networks (ANNs), support vector machines (SVMs) and linear regression (LR) with four feature selection methods, i.e., RFA, minimum redundancy maximum relevance (mRMR), support vector machine recursive feature elimination (SVM-RFE) and Relief. The performance of diagnosis of KBD with respect to different classification models were evaluated by sensitivity, specificity, accuracy, and the area under the receiver operating characteristic (ROC) curve (AUC).

RESULTS

Our results demonstrated that the 10 out of 26 discriminative features were displayed more powerful performance, regardless of the chosen of classification models and feature selection methods. These ten discriminative features were distal end of phalanges alterations, metaphysis alterations and carpals alterations and clinical manifestations of ankle joint movement limitation, enlarged finger joints, flexion of the distal part of fingers, elbow joint movement limitation, squatting limitation, deformed finger joints, wrist joint movement limitation.

CONCLUSIONS

The selected ten discriminative features could provide a fast, effective diagnostic standard for KBD adolescents.

Bayesian Gene Selection Based on Pathway Information and Network-Constrained Regularization

High-throughput data make it possible to study expression levels of thousands of genes simultaneously under a particular condition. However, only few of the genes are discriminatively expressed. How to identify these biomarkers precisely is significant for disease diagnosis, prognosis, and therapy. Many studies utilized pathway information to identify the biomarkers. However, most of these studies only incorporate the group information while the pathway structural information is ignored. In this paper, we proposed a Bayesian gene selection with a network-constrained regularization method, which can incorporate the pathway structural information as priors to perform gene selection. All the priors are conjugated; thus, the parameters can be estimated effectively through Gibbs sampling. We present the application of our method on 6 microarray datasets, comparing with Bayesian Lasso, Bayesian Elastic Net, and Bayesian Fused Lasso. The results show that our method performs better than other Bayesian methods and pathway structural information can improve the result.

Feature Selection Combining Information Theory View and Algebraic View in the Neighborhood Decision System

Feature selection is one of the core contents of rough set theory and application. Since the reduction ability and classification performance of many feature selection algorithms based on rough set theory and its extensions are not ideal, this paper proposes a feature selection algorithm that combines the information theory view and algebraic view in the neighborhood decision system. First, the neighborhood relationship in the neighborhood rough set model is used to retain the classification information of continuous data, to study some uncertainty measures of neighborhood information entropy. Second, to fully reflect the decision ability and classification performance of the neighborhood system, the neighborhood credibility and neighborhood coverage are defined and introduced into the neighborhood joint entropy. Third, a feature selection algorithm based on neighborhood joint entropy is designed, which improves the disadvantage that most feature selection algorithms only consider information theory definition or algebraic definition. Finally, experiments and statistical analyses on nine data sets prove that the algorithm can effectively select the optimal feature subset, and the selection result can maintain or improve the classification performance of the data set.

A fast and efficient count-based matrix factorization method for detecting cell types from single-cell RNAseq data

Background

Single-cell RNA sequencing (scRNAseq) data always involves various unwanted variables, which would be able to mask the true signal to identify cell-types. More efficient way of dealing with this issue is to extract low dimension information from high dimensional gene expression data to represent cell-type structure. In the past two years, several powerful matrix factorization tools were developed for scRNAseq data, such as NMF, ZIFA, pCMF and ZINB-WaVE. But the existing approaches either are unable to directly model the raw count of scRNAseq data or are really time-consuming when handling a large number of cells (e.g. n>500).

Results

In this paper, we developed a fast and efficient count-based matrix factorization method (single-cell negative binomial matrix factorization, scNBMF) based on the TensorFlow framework to infer the low dimensional structure of cell types. To make our method scalable, we conducted a series of experiments on three public scRNAseq data sets, brain, embryonic stem, and pancreatic islet. The experimental results show that scNBMF is more powerful to detect cell types and 10 - 100 folds faster than the scRNAseq bespoke tools.

Conclusions

In this paper, we proposed a fast and efficient count-based matrix factorization method, scNBMF, which is more powerful for detecting cell type purposes. A series of experiments were performed on three public scRNAseq data sets. The results show that scNBMF is a more powerful tool in large-scale scRNAseq data analysis. scNBMF was implemented in R and Python, and the source code are freely available at https://github.com/sqsun.

An Attribute Reduction Method Using Neighborhood Entropy Measures in Neighborhood Rough Sets

Attribute reduction as an important preprocessing step for data mining, and has become a hot research topic in rough set theory. Neighborhood rough set theory can overcome the shortcoming that classical rough set theory may lose some useful information in the process of discretization for continuous-valued data sets. In this paper, to improve the classification performance of complex data, a novel attribute reduction method using neighborhood entropy measures, combining algebra view with information view, in neighborhood rough sets is proposed, which has the ability of dealing with continuous data whilst maintaining the classification information of original attributes. First, to efficiently analyze the uncertainty of knowledge in neighborhood rough sets, by combining neighborhood approximate precision with neighborhood entropy, a new average neighborhood entropy, based on the strong complementarity between the algebra definition of attribute significance and the definition of information view, is presented. Then, a concept of decision neighborhood entropy is investigated for handling the uncertainty and noisiness of neighborhood decision systems, which integrates the credibility degree with the coverage degree of neighborhood decision systems to fully reflect the decision ability of attributes. Moreover, some of their properties are derived and the relationships among these measures are established, which helps to understand the essence of knowledge content and the uncertainty of neighborhood decision systems. Finally, a heuristic attribute reduction algorithm is proposed to improve the classification performance of complex data sets. The experimental results under an instance and several public data sets demonstrate that the proposed method is very effective for selecting the most relevant attributes with great classification performance.

Higher-order partial least squares for predicting gene expression levels from chromatin states

Background

Extensive studies have shown that gene expression levels are strongly affected by chromatin mark combinations via at least two mechanisms, i.e., activation or repression. But their combinatorial patterns are still unclear. To further understand the relationship between histone modifications and gene expression levels, here in this paper, we introduce a purely geometric higher-order representation, tensor (also called multidimensional array), which might borrow more unknown interactions in chromatin states to predicting gene expression levels.

Results

The prediction models were learned from regions around upstream 10k base pairs and downstream 10k base pairs of the transcriptional start sites (TSSs) on three species (i.e., Human, Rhesus Macaque, and Chimpanzee) with five histone modifications (i.e., H3K4me1, H3K4me3, H3K27ac, H3K27me3, and Pol II). Experimental results demonstrate that the proposed method is more powerful to predicting gene expression levels than several other popular methods. Specifically, our method enable to get more powerful performance on both commonly used criteria, R and RMSE, as high as 1.7% and 11%, respectively.

Conclusions

The overall aim of this work is to show that the higher-order representation is able to include more unknown interaction information between histone modifications across different species.

Ensemble based on static classifier selection for automated diagnosis of Mild Cognitive Impairment

BACKGROUND

Alzheimer's disease (AD) is the most common cause of neurodegenerative dementia in the elderly population. Scientific research is very active in the challenge of designing automated approaches to achieve an early and certain diagnosis. Recently an international competition among AD predictors has been organized: "A Machine learning neuroimaging challenge for automated diagnosis of Mild Cognitive Impairment" (MLNeCh). This competition is based on pre-processed sets of T1-weighted Magnetic Resonance Images (MRI) to be classified in four categories: stable AD, individuals with MCI who converted to AD, individuals with MCI who did not convert to AD and healthy controls.

NEW METHOD

In this work, we propose a method to perform early diagnosis of AD, which is evaluated on MLNeCh dataset. Since the automatic classification of AD is based on the use of feature vectors of high dimensionality, different techniques of feature selection/reduction are compared in order to avoid the curse-of-dimensionality problem, then the classification method is obtained as the combination of Support Vector Machines trained using different clusters of data extracted from the whole training set.

RESULTS

The multi-classifier approach proposed in this work outperforms all the stand-alone method tested in our experiments. The final ensemble is based on a set of classifiers, each trained on a different cluster of the training data. The proposed ensemble has the great advantage of performing well using a very reduced version of the data (the reduction factor is more than 90%). The MATLAB code for the ensemble of classifiers will be publicly available¹ to other researchers for future comparisons.

A high-order representation and classification method for transcription factor binding sites recognition in Escherichia coli

BACKGROUND

Identifying transcription factors binding sites (TFBSs) plays an important role in understanding gene regulatory processes. The underlying mechanism of the specific binding for transcription factors (TFs) is still poorly understood. Previous machine learning-based approaches to identifying TFBSs commonly map a known TFBS to a one-dimensional vector using its physicochemical properties. However, when the dimension-sample rate is large (i.e., number of dimensions/number of samples), concatenating different physicochemical properties to a one-dimensional vector not only is likely to lose some structural information, but also poses significant challenges to recognition methods.

MATERIALS AND METHOD

In this paper, we introduce a purely geometric representation method, tensor (also called multidimensional array), to represent TFs using their physicochemical properties. Accompanying the multidimensional array representation, we also develop a tensor-based recognition method, tensor partial least squares classifier (abbreviated as TPLSC). Intuitively, multidimensional arrays enable borrowing more information than one-dimensional arrays. The performance of each method is evaluated by average F-measure on 51 Escherichia coli TFs from RegulonDB database.

RESULTS

In our first experiment, the results show that multiple nucleotide properties can obtain more power than dinucleotide properties. In the second experiment, the results demonstrate that our method can gain increased prediction power, roughly 33% improvements more than the best result from existing methods.

CONCLUSION

The representation method for TFs is an important step in TFBSs recognition. We illustrate the benefits of this representation on real data application via a series of experiments. This method can gain further insights into the mechanism of TF binding and be of great use for metabolic engineering applications.