Comparative analysis of Microarray and RNAseq data from liver cancer samples

Abstract

Background: The field of transcriptomics has developed an exponential growth in the last years thanks to the advances in high throughput RNA sequencing (RNAseq), which is becoming more popular as the cost is decreasing. Traditional techniques to study transcriptomics like microarrays are trying to survive in this new scenario by developing new microarrays. New microarray chips like the Human Transcriptome Array 2.0 (HTA2.0, Affymetrix) are nowadays co-existing with RNAseq but the performance of these new arrays have not been compared yet to RNAseq. Results: Here we show a comparative analysis of RNAseq and HTA2.0 in terms of gene expression. We observe that there is a good concordance between the two techniques, but there are some differences that may be considered when choosing one. Using hepatoblastoma samples, we have worked at 3 levels. The first approach was to study different methods to normalize microarray data. We found that using SST-RMA normalization was the best method to examine HTA2.0 data. At a second level, we have studied different ways to analyze RNAseq data. We have worked with two splice aware aligners, HISAT2 and STAR, and we found that despite there are genes that have a significant difference in counts, this does not affect subsequent steps. Furthermore, we use 5 different tools to assess gene expression genes from count data. We have found that tweeDEseq gave us the best results when analyzing RNAseq data. Finally, we have compared microarray against RNAseq results, and we have found that despite having some differences, there is a good concordance between both techniques, in terms of finding differentially expressed genes and at functional level. Conclusions: Both RNAseq and microarrays are good options to study transcriptomics, as they give similar results at gene expression level. RNAseq has the advantage that more analysis can be performed, like variant calling, alternative splicing or finding new transcripts. On the other side, HTA2.0 compete with a more affordable cost, and an easier and standardized data analysis. Furthermore, HTA2.0 has probes to detect low expressed transcripts or short transcripts, which are lost in most RNAseq analyses. HTA2.0 can also be used to study alternative splicing, so is an interesting option if the researcher wants a relatively fast analysis of transcriptomics.