Resumen de Computational analysis and characterization of alternative splicing and its impact on transcriptional diversity
In all eukaryotes, pre-mRNA splicing is a requirement for protein synthesis as most coding sequences are interspersed with non-coding ones that need to be removed from the primary transcript. The basic steps of this process are stereotyped and essentially universal but, as nucleotide sequences can vary greatly and in an unpredictable fashion, a complex mechanism of regulation ensures the efficient and flexible processing of myriads of different substrates. By virtue of this flexibility, the same pre-mRNA can be spliced in different ways and alternative isoforms are thereby generated from a single genomic template. This alternative splicing is essentially a by-product of the process, but can profoundly affect the transcriptional and protein-coding potential of eukaryotic genes. Its global impact remains however elusive and the functional relevance of alternatively spliced transcripts is a controversial and open-ended question.
The advent of RNA sequencing enabled the investigation of splicing patterns at a genome-wide scale. Best analysis practices are nonetheless difficult to establish, as the alignment of short sequence fragments is a complex computational task. This task is made even more challenging by splicing itself, as different parts of these fragments originate from separate genomic locations and are harder to map unambiguously. In order to tackle this issue, simulated sequencing data was generated to evaluate systematically the effectiveness of different alignment algorithms in terms of splice site mapping, detection and quantification. As none of these algorithms provided an integrated solution to the three problems, a novel computational strategy was devised to achieve a better trade-off in performance. The tool that achieved best mapping and quantification accuracy was coupled to a newly implemented method for efficient splice site detection. This pipeline was tested in both simulated and real data, proving to achieve consistently better results.
Following implementation and testing, the established strategy was used to analyse comparative transcriptomics data in different tissues. Tissue-dependent splicing signatures were identified and one-to-one orthologs were found to share similar patterns in primates and mouse. Though rare, instances of tissue-specific splicing were prevalently observed in genes expressed at constitutively high levels and with more strongly constrained, CpG-rich, promoters. Despite being ubiquitously expressed, these genes exhibited enhanced mRNA levels in the specific tissue where alternative splicing occurs. This was further associated with an increase in protein content whenever the splicing change was found to be common to both human and mouse. Modulation of tissue-specific abundances in proteins encoded by constitutively high expressed genes provides a rationale for conservation of alternative splicing patterns.
