Fig. 3
From: Characterizing RNA stability genome-wide through combined analysis of PRO-seq and RNA-seq data
Features of transcription units (TUs) that are predictive of transcription rate and RNA half-life. a Structural equation model (SEM) describing the effects of an arbitrary collection of TU features (X1, …, XN, with intercept term X0 = 1) on transcription rate (b) and half-life (t1/2), as well as the downstream impact on mRNA concentration (m), normalized PRO-seq (p), and normalized RNA-seq (r) read counts. The model is linear in logarithmic space, with unmodeled variation accounted for as Gaussian noise (ɛb, ɛt, ɛp, and ɛr; see the “Methods” section). The coefficients for transcription rate (λn) and half-life (μn) are estimated by maximum likelihood, assuming independence of replicates and pooling data from all TUs of the same class. b Estimated values for coefficients for transcription (λn; top) and half-life (μn; bottom) for various features of interest. Results are for intron-containing mRNAs (see Additional file 2: Figures S13 & S14 for other classes). Features considered for each TU: G+C 3′UTR, GC content in 3′UTR; G+C 5′UTR, GC content in 5′UTR; G+C cds, GC content in coding region; G+C intron, GC content in intron(s); len 3′UTR, length of 3′UTR; len 5′UTR, length of 5′UTR; len cds, total length of coding region; len intron, total length of intron(s); spl. junc. dens., number of splice junctions divided by mature RNA length. Error bars represent ± 1.96 standard error, as calculated by the “lavaan” R package [39]. Significance (from Z-score): *p < 0.05; **p < 0.005; ***p < 0.0005