Fig. 3.

Modeling background distribution of TI-seq data. a An illustration of the typical TI-seq RPF count profile across a hypothetical protein-coding transcript. The RPF counts at the first base of CDS in-frame codons (the pink bars inside ORF), excluding AUG or near-cognate start codons, from annotated PCGs, were used to model TI-seq data background. TI-seq RPF count profile for the major isoform of b GAPDH and c UBTD1. d Fitting of different distributions including Poisson, zero-inflated Poisson (ZIP), negative bionomial (NB), and zero-inflated negative binomial (ZINB) to the observed background RPF count distribution. e NB distribution parameters (r and p) estimated from different TI-seq expression groups. f The use of different NB background distributions for transcripts/genes with different TI-seq signal density improved TIS identification compared to the use of a single/global NB background distribution