Robust selection of cancer survival signatures from high-throughput genomic data using two-fold subsampling.

Sangkyun Lee,Melanie Schwermer,Rogier Versteeg,Jörg Rahnenführer,Shahab Asgharzadeh,Raymond L Stallings,Kathrin Fielitz,Luigi Varesio,Jan Koster,Pieter Mestdagh,Michel Lang,Katleen De Preter,Johannes H Schulte,Alexander Schramm,Katharina Morik,Ioannis P Androulakis

doi:10.1371/journal.pone.0108818

Abstract

Identifying relevant signatures for clinical patient outcome is a fundamental task in high-throughput studies. Signatures, composed of features such as mRNAs, miRNAs, SNPs or other molecular variables, are often non-overlapping, even though they have been identified from similar experiments considering samples with the same type of disease. The lack of a consensus is mostly due to the fact that sample sizes are far smaller than the numbers of candidate features to be considered, and therefore signature selection suffers from large variation. We propose a robust signature selection method that enhances the selection stability of penalized regression algorithms for predicting survival risk. Our method is based on an aggregation of multiple, possibly unstable, signatures obtained with the preconditioned lasso algorithm applied to random (internal) subsamples of a given cohort data, where the aggregated signature is shrunken by a simple thresholding strategy. The resulting method, RS-PL, is conceptually simple and easy to apply, relying on parameters automatically tuned by cross validation. Robust signature selection using RS-PL operates within an (external) subsampling framework to estimate the selection probabilities of features in multiple trials of RS-PL. These probabilities are used for identifying reliable features to be included in a signature. Our method was evaluated on microarray data sets from neuroblastoma, lung adenocarcinoma, and breast cancer patients, extracting robust and relevant signatures for predicting survival risk. Signatures obtained by our method achieved high prediction performance and robustness, consistently over the three data sets. Genes with high selection probability in our robust signatures have been reported as cancer-relevant. The ordering of predictor coefficients associated with signatures was well-preserved across multiple trials of RS-PL, demonstrating the capability of our method for identifying a transferable consensus signature. The software is available as an R package rsig at CRAN (http://cran.r-project.org).

Highlights

Identification of relevant features from large data sets has been a focus of many research fields for a long time
When sample sizes are small as in most clinical studies, such practices can lead to identifying diverse signatures from multiple studies that look perfectly fine on their own evaluation but are not successful when they are applied to the data from other studies
Our robust selection (RS) framework successfully improved the robustness of the popular multivariate signature selection methods, the lasso (L) and the preconditioned lasso (PL) algorithms, for predicting survival risk: this was the primary goal of this paper

Summary

Introduction

Identification of relevant features from large data sets has been a focus of many research fields for a long time. Robustness is a critical factor especially in clinical studies, when the purpose is either to identify the key players in the underlying biological systems, or to develop clinically useful tests. A typical example is to perform feature selection on a single partition of available cohort data, to determine the success of selection using the rest of data (often called as a test set). When sample sizes are small as in most clinical studies, such practices can lead to identifying diverse signatures from multiple studies that look perfectly fine on their own evaluation but are not successful when they are applied to the data from other studies

Methods

Results

Conclusion