Fig. 2

Calling transcription units from GRO-seq data using groHMM. a Schematic representation of the groHMM hidden-Markov model approach. The emission probabilities of each state (i.e., transcribed and non-transcribed) were modeled with gamma distributions. Red arrows represent two reserved tuning parameters for the model; T, the transition probability of the transcribed state to the non-transcribed state and σ² _, the variance of the non-transcribed state in a constrained gamma distribution. Γ(σ², 1/σ²), constrained gamma distribution of the non-transcribed state; Γ(k_T, θ_T), gamma distribution of the transcribed state; N, the transition probability of the non-transcribed to the transcribed state. Gray arrows, self-transition probabilities (i.e., transcribed to transcribed or non-transcribed to non-transcribed), which are, by definition, 1-T and 1-N, respectively. b Genome browser tracks of GRO-seq data from MCF-7 cells (top) with transcription units called by groHMM, SICER, and HOMER (middle), and corresponding RefSeq annotations (bottom). c Zoomed in view of the browser tracks for the RPS6KC1 gene from (b)