Elastic net-based prediction of IFN-β treatment response of patients with multiple sclerosis using time series microarray gene expression profiles

Abstract

INF-β has been widely used to treat patients with multiple sclerosis (MS) in relapse. Accurate prediction of treatment response is important for effective personalization of treatment. Microarray data have been frequently used to discover new genes and to predict treatment responses. However, conventional analytical methods suffer from three difficulties: high-dimensionality of datasets; high degree of multi-collinearity; and achieving gene identification in time-course data. The use of Elastic net, a sparse modelling method, would decrease the first two issues; however, Elastic net is currently unable to solve these three issues simultaneously. Here, we improved Elastic net to accommodate time-course data analyses. Numerical experiments were conducted using two time-course microarray datasets derived from peripheral blood mononuclear cells collected from patients with MS. The proposed methods successfully identified genes showing a high predictive ability for INF-β treatment response. Bootstrap sampling resulted in an 81% and 78% accuracy for each dataset, which was significantly higher than the 71% and 73% accuracy obtained using conventional methods. Our methods selected genes showing consistent differentiation throughout all time-courses. These genes are expected to provide new predictive biomarkers that can influence INF-β treatment for MS patients.

Type I Interferon (IFN)-Regulated Activation of Canonical and Non-Canonical Signaling Pathways

“…One study used a feature selection computational method on a longitudinal microarray dataset of relapse-remitting MS (RRMS) patients treated with IFNb-1b, and found a predictive seven gene signature (CXCL9, IL2RA, CXCR3, AKT1, CSF2, IL2RB, GCA) with 65.08% predictive accuracy (59). Using an alternative method of Elastic net modeling, Fukushima et al analyzed time-course microarray datasets from PBMCs of MS patients and identified eleven (ZBTB16, ZFP37, HPS5, HOPX, ARFGAP3, CALML5, VPS26A, SLC5A4, MBL2, DLGAP4, CACNA1C) and eight (SMA4, MIR7114_NSMF, LSM8, FLAD1, RRN3P1, RASL10A, IER3IP1, CDH2) genes predictive of an IFNb response, with 81% and 78% accuracy, respectively, for each dataset (60). A different study employed the GeneRank method to identify monotonically expressed genes (MEGs) that determine a good response (AFTPH, ALOX5, ATG7, MYD88, LILRB1, PRKAB1, PSEN1, VAMP3) and a bad response (AGFG1, CHM, IGLL1, PELI1, PTEN) for responders, and two bad response MEGs for non-responders (NAP1L4, MMS19) in IFNb treated RRMS patients (61).…”

A recursive framework for predicting the time-course of drug sensitivity

“…We examine the predictive ability on the prediction of binary drug response and ordinal EDSS response. For the binary case, we apply a number of classifiers including two linear models (EN-LR 8 and SVM), one nonlinear model ( K -nearest neighbors (KNN) 30 ), and a probabilistic graphical model (discriminative loop hidden Markov model (dl-HMM) 14 ) to real-world time-course data. We did not include SVM with nonlinear kernels (e.g.…”

“…In parallel with these technological developments, there has been growing interest in the application of machine learning methods to analyze the time-course gene expression data. For example, time-course gene expression data can be used to not only identify longitudinal phenotypic markers 11 , 12 , but also assess the association between the time course molecular data and cytokine production in this HIV trial 13 and predict drug response during a treatment 8 , 14 . In 15 , the authors proposed an integrated Bayesian inference system to select genes for drug response classification from time-course gene expression data.…”

“…Recently, Fukushima et al 8 developed an algorithm for joint gene selection and drug response prediction for time-course data. The method uses Elastic-Net (EN) to select a set of genes that show discrimination of patients’ drug responses throughout the treatment.…”

Identification of Monotonically Differentially Expressed Genes for IFN-β-Treated Multiple Sclerosis Patients

“…While the therapeutic mechanism of IFN-β has not been fully elucidated, microarray experiments have been used to evaluate long-term response to IFN-β treatment. However, the resulting longitudinal gene expression profiles have usually been analyzed using statistical methods incapable of dealing with such data [5]. e inconsistency between the data and the analysis methods may cause biased or even totally incorrect conclusions, making it difficult to unravel the mechanism of action of IFN-β in MS. erefore, a reanalysis of longitudinal gene expression data using a machine learning method capable of identifying genes that present a consistently changed pattern across time is recommended.…”