Improved funnel-gsea using adaptive elastic-net penalization method to identify significant gene sets

Abstract

Gene set enrichment analysis (GSEA) is one of the methods in functional class scoring (FCS) categories for gene set analysis. GSEA is a popular method that was developed to identify, analyse and interpret set of genes or pathways from high-throughput transcriptomics experiments which are significantly enriched to help further analysis by biologist researchers. Many methods have been developed to enhance the original procedure of the GSEA. One of the evolutions of the GSEA method is the use of the elastic-net to reduce the effect of overlapping that reduces the statistical power and instability of the inference at the level of the gene set. However, elastic-net has limitations as it is inconsistent and bias in estimation. Thus, an ADaptive ELastic-NET in GSEA (ADELNET-GSEA) with an adaptive elastic-net was proposed to achieve a better result in identifying more gene sets that are informative and significant. The key part of the adaptive elastic-net is the weight parameter. It enables the adaptive elastic-net to perform different amounts of shrinkage to the different variables. Consequently, the ADELNET-GSEA is also consistent and unbiased in estimation. This research utilized the real dataset of Influenza A H3N2 time-course gene expression. It was found that the ADELNET-GSEA outperformed the previous GSEA method by identifying higher numbers of informative and significant gene sets to the immune response to human influenza infection. ADELNET-GSEA was able to identify the new gene sets, which were Spliceosome and Ubiquitin Mediated Proteolysis gene sets, related to the immune response for influenza. These findings have been validated through a word search strategy proven by previous researchers. Based on this result, this research brings benefits to the biological context validation and able to clarify the reliability of the improved method in identifying the significant gene sets.