McGurk, M. P.; McWatters, D. C.; Burge, C. B.

It is well-established that splicing factors (SFs) regulate splicing in a position-dependent manner. However, despite extensive experimental work focused on this regulation, quantifying SF regulatory activity in a form that can be readily applied to new analyses remains a challenge. Here we present an interpretable regression model, KATMAP, which explicitly models splicing changes upon SF knockdown in terms of changes in SF binding across the transcriptome and the resulting loss of regulation. This allows us to learn an SF\'s positional activity and predict the factor\'s regulatory impact on individual exons, yielding regulatory models consistent with prior literature and highly reproducible across knockdowns of the same SF in different cell types, even generalizing well to related species. We leverage this generalizability to distinguish direct targets from the indirect effects of knockdown, and identify the secondary SFs responsible for these widespread indirect effects. We further uncover evidence of cooperative regulation between RBFOX and QKI. The generality of our statistical framework means it could be extended to the study of RNA regulation and processing beyond splicing.