The Influence of Feature Selection Methods on Accuracy, Stability and Interpretability of Molecular Signatures

By Anne-Claire Haury, Pierre Gestraud, Jean-Philippe Vert. PLoS ONE, 2011

400 citations recorded by the source corpus.

Biomarker discovery from high-dimensional data is a crucial problem with enormous applications in biology and medicine. It is also extremely challenging from a statistical viewpoint, but surprisingly few studies have investigated the relative strengths and weaknesses of the plethora of existing feature selection methods. In this study we compare feature selection methods on public gene expression datasets for breast cancer prognosis, in terms of predictive performance, stability and functional interpretability of the signatures they produce. We observe that the feature selection method has a significant influence on the accuracy, stability and interpretability of signatures. Surprisingly, complex wrapper and embedded methods generally do not outperform simple univariate feature selection methods, and ensemble feature selection has generally no positive effect. Overall a simple Student's t-test seems to provide the best results.

Subfields: Molecular Biology

View Paper: https://www.semanticscholar.org/paper/5f9a5c6925886b3789d7e702531521b490aaaebb

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

Experimental analysis of new algorithms for learning ternary classifiers (2015, background): “Learning classifiers in very high dimension data has received more and more attention in the past ten years both theoretically [2], [3], [4], [5], [6], [7] and in practice through Data Mining applications [8], [9], [10], [11], [12], [13], [14], [15], [16], [17].”
An experimental study of the intrinsic stability of random forest variable importance measures (2016, neutral): “, 10-fold cross validation) [15, 19–21].”
Machine Learning methods for Quantitative Radiomic Biomarkers (2015, methodology): “In order to assess the stability of feature selection methods, we used a stability measure proposed by Yu et al. 32 under the hard data perturbation settings 33 .”
Feature Selection Based on Structured Sparsity: A Comprehensive Study (2017, background): “To solve this issue, feature selection (also known as feature ranking, subset, or variable selection) [1]–[7] techniques are designed to select a subset of features from the high-dimensional feature set for a compact and accurate data representation.”
Deep learning in cancer diagnosis, prognosis and treatment selection (2021, background): “Standard computational methods, such as support vector machines (SVMs) or k-nearest neighbours, used to subtype cancers can be prone to errors due to batch effects [90] and may rely only on a handful of signature genes, omit-ting important biological information [91 – 93].”
A survey on feature selection methods (2014, methodology): “Ensemble feature selection [68,69] is a relatively new technique used to obtain a stable feature subset.”
Methods and Models for Interpretable Linear Classification (2014, background): “These domains include credit scoring [76], crime prediction [102, 2, 93], national defense [27], marketing [49, 114], medical diagnosis [106, 112], and scientific discovery [104, 36, 48].”
Concise Polygenic Models for Cancer-Specific Identification of Drug-Sensitive Tumors from Their Multi-Omics Profiles (2020, methodology): “We opted to use univariate filter methods to avoid long computational times, making assumptions or biasing predictive performance, as previously reported in the literature [28] (this study found univariate filters to be among the top-performing feature selection algorithms in a related highdimension”
A feature selection model based on genetic rank aggregation for text sentiment classification (2017, methodology): “For each list, whether the examined feature exceeds the threshold value or not is determined and based on this judgement, a position is assigned to the feature [48].”
Correlation and variable importance in random forests (2013, methodology): “Haury et al. [16] provide a comparison of ensemble feature selections combined with several classiﬁcation methods.”