Minimum redundancy feature selection from microarray gene expression data

By Chris H. Q. Ding, Hanchuan Peng. Computational Systems Bioinformatics. CSB2003. Proceedings of the 2003 IEEE Bioinformatics Conference. CSB2003, 2003

2,985 citations recorded by the source corpus.

Selecting a small subset of genes out of the thousands of genes in microarray data is important for accurate classification of phenotypes. Widely used methods typically rank genes according to their differential expressions among phenotypes and pick the top-ranked genes. We observe that feature sets so obtained have certain redundancy and study methods to minimize it. Feature sets obtained through the minimum redundancy - maximum relevance framework represent broader spectrum of characteristics of phenotypes than those obtained through standard ranking methods; they are more robust, generalize well to unseen data, and lead to significantly improved classifications in extensive experiments on 5 gene expressions data sets.

Subfields: Molecular Biology

View Paper: https://www.semanticscholar.org/paper/86d2df70fa9caddad48f479a902b8e052852ac51

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Molecular Biology

Citations in the corpus

SVM-RFE With MRMR Filter for Gene Selection (2010, methodology): “We used MATLAB toolbox from the original authors for experiments with MRMR method [6].”
Information-Theoretic Feature Selection in Microarray Data Using Variable Complementarity (2008, methodology): “For these reasons we recently assisted in filter literature to the adoption of selection criteria based on bi- and tri-variate densities at most [14], [17], [46].”
A review of microarray datasets and applied feature selection methods (2014, methodology): “Apart from the application of multivariate filter methods [34,126,121,45], the microarray problem was also treated with more complex techniques such as wrappers and embedded methods [22,63,60,97].”
Feature Extraction and Selection for Emotion Recognition from EEG (2014, methodology): “The most famous method using mutual information to characterize the suitability of a feature subset is called minimalRedundancy-Maximal-Relevance (mRMR) developed by Ding and Peng [52].”
Genetic programming for feature construction and selection in classification on high-dimensional data (2015, background): “All datasets are discretised in the same way as in [9] to reduce noise.”
Multimodal deep learning models for early detection of Alzheimer’s disease stage (2021, methodology): “…(mRMR) 49 We chose mRMR as a feature selec-Figure tion method because it works well with categorical data (such the SNP data) and has been previously reported with genetic data 50 . mRMR was chosen over other wrapper-based techniques such as sequential feature selection due to computational costs.”
Feature selection based on mutual information criteria of max-dependency, max-relevance, and min-redundancy (2003, background): “Some researchers have studied indirect or direct means to reduce the redundancy among features 1 (e.g., [4], [14], [19], [15], [22], [12], [5]) and select features with the minimal redundancy (Min-Redundancy).”
A Review of Feature Selection and Feature Extraction Methods Applied on Microarray Data (2015, methodology): “…is used to find the maximum relevance between a gene and a class label and then the correlation of the gene pair in that class is measured to minimise redundancy [15] as shown in (3) and (4), respectively: where 𝐹 is the 𝐹 -statistic, 𝑖 and 𝑗 are genes, ℎ is a class label, |𝑆| is the number…”
minet: A R/Bioconductor Package for Inferring Large Transcriptional Networks Using Mutual Information (2008, background): “It has been shown in [19] that the MRMR criterion is an optimal "pairwise" approximation of the conditional mutual information between any two genes X i and X j given the set S of selected variables I ( X i ; X j | S ).”
Incremental wrapper-based gene selection from microarray data for cancer classification (2006, neutral): “Other methods based on relevance and redundancy concepts can be found in [5,8].”
Data-analysis strategies for image-based cell profiling (2017, background): “Alternatively, if all concentrations are expected to have a phenotype, an entire compound can be left out of the analysis when its doses do not cluster together consistently 107 .”
Predicting cancer outcomes with radiomics and artificial intelligence in radiology (2021, background): “Features can be chosen to optimize predictive performance 32,33 , reduce correlation within a feature set 34 or maximize robustness and stability 35 .”