Comparison of genome-wide gene expression profiling by RNA Sequencing versus microarray in bronchial biopsies of COPD patients before and after inhaled corticosteroid treatment: does it provide new insights?

Abstract

In the era of “big data”, microarray technology has provided researchers with the ability to measure the expression of thousands of genes in a single experiment [1]. However, array technology is limited, as it can only measure transcripts present in medium to high abundance and can only quantify genes for which oligonucleotide probes are specifically designed. RNA-Seq, the direct sequencing of RNA, is rapidly becoming more popular in analysing gene expression. RNA-Seq performs better with respect to the detection of low-abundance transcripts, identifying genetic variants and detecting more differentially expressed genes with higher fold-change [2, 3]. Bulk tissue cell-type deconvolution represents a recently developed computational method to interrogate the proportions of cell types in a sample using cell type specific gene expression references [4]. This method is mainly based on RNA-Seq data; however, little has been done to determine whether this technique can be utilised for microarray technology. We sought to investigate whether gene expression profiling in COPD bronchial biopsies, using RNA-Seq, provides additional insight into the transcriptional effects before and after inhaled corticosteroids (ICS), compared to microarrays. Furthermore, we aimed to determine whether cellular deconvolution techniques can be conducted on microarray data by using two current methods: non-negative least squares (NNLS) and support vector regression (SVR), which tries to fit the regression within a certain threshold, and comparing them to RNA-Seq data. To this end, we analysed the steroid response before and after 6\u2005months of ICS treatment in participants with COPD. Therefore, we utilised gene expression data from bronchial biopsies, which were measured using both microarray (Affymetrix Hugene_ST1.0 array) and RNA-Seq (Illumina HiSeq 2500 platform). The bronchial biopsies were obtained from the Groningen and Leiden Universities Study of Corticosteroids in Obstructive Lung Disease (GLUCOLD) [5]. The methods of microarray sequencing in GLUCOLD have been described previously [6]. With respect to RNA-Seq, the RiboZero GOLD libraries were sequenced using 50\u2005bp single-read sequencing. The FastQC programme (version 0.11.5; https://github.com/s-andrews/FastQC) was utilised to perform quality control checks on the raw sequence data; the sequences were then trimmed using the java programme trimmomatic 0.33 [7]. The RNA-Seq mapping was conducted using Spliced Transcripts Alignment to a Reference (STAR) version 2.5.3a [8]. Principal component analysis was performed (using R) to detect extreme outliers. After these quality checks, all samples were found to be of sufficient quality. More DEGs are detected by RNA-Seq than microarrays in COPD lung biopsies and are associated with immunological pathways. Performing bulk tissue cell-type deconvolution in microarray lung samples, using the SVR method, reflects RNA-Seq results. https://bit.ly/2N8sY3s