Quantifying Transcriptional Rates with MPRAnalyze

The quantification of induced transcriptional rates (“alpha” values) was performed using MPRAnalyze [29 (link)]. Briefly, MPRAnalyze fits two nested generalized linear models (GLMs): the first estimates the latent construct counts from the observed DNA counts, and the second estimates the latent rate of transcription from the latent construct estimates and observed RNA counts. The models are optimized using likelihood maximization, with a gamma likelihood for the DNA counts and a negative binomial likelihood for the RNA counts. MPRAnalyze includes library-size normalization factors, which were computed once using the entire dataset and then used across all analyses, to maintain consistency. For the quantification of alpha values, the full experimental design was included in the design matrix for the DNA model, and an alpha value was extracted for each replicate RNA model.