Predicting human mRNA isoform levels from site-specific splicing kinetics in silico

Splicing of pre-mRNA can result in multiple possible mRNA isoforms per gene due to alternative splicing. The frequency at which individual isoforms occur depends on the intrinsic splicing kinetics of the pre-mRNA as well as intracellular chemical conditions. Computational modeling can potentially provide a platform to rapidly assess how variations in intracellular and environmental conditions, for example differential levels of regulatory splicing proteins, affect kinetics and resulting mRNA isoforms. Overcoming the vast combinatoric possibilities of splicing, however, has remained a significant challenge in modeling its kinetics. Here we report the development of a stochastic kinetic model of splicing that is extensible to most protein-coding genes in the human genome. Our model allows for variations in site-specific reaction rates as well as the ability to introduce additional splicing factors. We experimentally validate the predictive capability of our computational model by exploring the spliced isoform ratio of a target gene (SRSF6) under normoxia and hypoxia. This work provides a resource for quantitative, computational analysis of pre-mRNA splicing, allowing for a rapid computational-experimental approach to assess biological hypotheses.