One of the main challenges in modern biology is understanding how and when genes are turned on. As our knowledge of transcription regulation has matured, bioinformatic analyses have allowed increasingly quantitative predictions of gene expression. The ultimate goal of such analyses is to predict gene expression from concentrations of proteins in the cell, based on protein-DNA networks. Yet this is a highly ambitious task in Eukaryotes, since their gene expression is determined by chromatin conformation, epigenetic factors, promoter and enhancer states, rates of transcription initiation, elongation, transcript processing and termination, and mRNA export and stability. This thesis is focused on the co-occurrence of transcriptional elongation and pre-mRNA splicing, the process in which introns are removed from the pre-mRNA transcript. Splicing is an important regulatory step because aberrant splicing leads to either reduced or non-functional protein expression, and alternative splicing expands the repertoire of functional proteins encoded by the genome. The co-transcriptional nature of splicing implies that the kinetics of elongation in relation to splicing are important for the outcome of splicing decisions. Co-transcriptional splicing (CTS) has been extensively studied, but quantitative models of transcription networks that predict gene expression timecourses have yet to incorporate CTS considerations. Here I constructed kinetic models of CTS. Initially, I built a model of constitutive CTS and developed methods to fit nascent RNA-seq data to the model. Fitting this model to published datasets indicated that only a subset of genes can be expected to process all of their introns co-transcriptionally. Detailed data-mining of high-throughput datasets and genomes revealed patterns of compensatory signatures in sequence, chromatin and polymerase data, suggesting an evolutionary selection towards splicing co-transcriptionally. Next I expanded the model to include alternative splicing reactions. Despite the exponential combinatorial complexity, all possible isoforms resulting from up to nine introns can be simulated. A further expansion of the model considers separate reactions at the 5' and 3' ends of introns, which allows for simulation of phenomena such as exon definition and polymerase-mediated recruitment. Together, these novel tools can be used to test quantitative predictions of genome-wide splicing outcomes, or be incorporated into larger gene expression models.