Identification of Differential Alternative Splicing Events: Assessing Tools Performance with Different Sequencing Parameters

Alternative splicing (AS) is a crucial process in transcript maturation, enabling the generation of multiple transcript variants and the control of transcript abundance. In gene expression analysis, accounting for AS is crucial as different transcripts, arising from the same gene, encode distinct proteins. Moreover, the identification of differences in the proportions of AS events enrich the classical differential expression analysis with additional insight regarding cell behaviour. Most emerging AS tools focus on individual splicing events and are able to quantify the proportion of AS events from RNA-seq data. The systematic evaluation of such tools is lacking as: i) proposed benchmark studies do not consider the distinction between various types of AS events, neglecting the different performances that a tool may have on different events; ii) the proposed benchmark studies adopt single sample RNA-seq data and do not consider the potential impact of sequencing parameters on tools’ performance; and iii) confounding factors such as gene expression level may be taken into consideration as they may affect tools performance. We propose an extensive benchmark study to evaluate the performance of three widely used event-based AS tools: SUPPA, rMATS, and MAJIQ, selected based on their citations. Moreover, we propose a novel reusable benchmark dataset generator to create gold-standards from real RNA-seq data. We show that SUPPA consistently outperforms others, followed by rMATS and MAJIQ. Performance were evaluated by varying sequencing parameters, such as library size, fragment length, read length and sequencing type (paired-end or single-end). Results show that SUPPA’s performance remains generally stable regardless of and is not affected by sequencing parameters. In contrast, the performance of rMATS and MAJIQ is moderately affected by sequencing parameters.