bioRxiv2023
Genome-wide expression data has become ubiquitous within the last two decades. Given such data, functional enrichment methods identify functional categories (e.g., biological processes) that preferentially annotate differentially expressed genes. However, many existing methods operate in a binary manner, disregarding valuable information contained in the gene ranking. The few methods that consider the ranking often return redundant or non-specific functional categories. To address these limitations, we developed a novel method called Concise Ranked Functional Enrichment (CRFE), which effectively leverages the ranking information in gene expression data to compute a non-redundant set of specific functional categories that are notably enriched for highly ranked genes. A particularly useful feature of CRFE is a tunable parameter that defines how much focus should be given to the most highly ranked genes. Using four treatment-control RNA-seq datasets, we compared the performance of CRFE with the two most widely used types of functional enrichment methods, Gene Set Enrichment Analysis and over-representation analysis. We evaluated the methods based on their ability to utilize ranking information, generate non-redundant results, and return functional categories with high information content. CRFE excelled in all evaluated criteria, outperforming the existing methods, each of which exhibits deficiencies in at least one aspect. Using lung adenocarcinoma data, we further showed that the functional categories identified by CRFE are biologically meaningful. In conclusion, CRFE computes an informative set of functional categories that summarizes genome-wide expression data. With its superior performance over existing methods, CRFE harbors great promise to become a widely used functional enrichment method. Author summaryGiven a list of differentially expressed genes as input, functional enrichment methods reveal which functional categories (e.g., biological processes) were likely activated by the cell and are responsible for the differential expression. We developed a new such method, called Concise Ranked Functional Enrichment (CRFE), which addresses the limitations of current approaches by incorporating gene ranking information to compute a concise and specific set of enriched functional categories. Using four treatment-control RNA-seq datasets, we evaluate how well CRFE and the two currently most widely used methods perform in three criteria. We find that CRFE outperforms each of the alternative methods in at least one of the evaluated criteria, demonstrating its superiority. A high-level interpretation of the functional categories identified by CRFE for lung adenocarcinoma datasets highlights its usefulness for experimentalists. Overall, CRFE harnesses the power of ranked gene lists to generate a focused and non-redundant set of enriched functional categories. Our study positions CRFE as a promising method for functional enrichment analysis, with the potential to advance research in this field.