Accurate design of translational output by a neural network model of ribosome distribution

Abstract

Synonymous codon choice can have dramatic effects on ribosome speed and protein expression. Ribosome profiling experiments have underscored that ribosomes do not move uniformly along mRNAs. We modeled this variation in translation elongation using a feedforward neural network to predict the ribosome density at each codon as a function of its sequence neighborhood. Our approach revealed sequence features affecting translation elongation and characterized large technical biases in ribosome profiling. We applied our model to design synonymous variants of a fluorescent protein spanning the range of translation speeds predicted with our model. Levels of the fluorescent protein in budding yeast closely tracked the predicted translation speeds across their full range. We therefore demonstrate that our model captures information determining translation dynamics in vivo , that we can harness this information to design coding sequences, and that control of translation elongation alone is sufficient to produce large, quantitative differences in protein output.

Analysis of computational codon usage models and their association with translationally slow codons

“…17 was arbitrarily 144 chosen as a compromise between smaller windows that were relatively noisy, and larger 145 windows that could dilute an individual codon's contribution. Another study [2], which 146 aims to predict RFP counts (and therefore local translation rates), found that a window 147 of (-5, +4) around the A-site (i.e., 5 codons to the left of the A-site, the A-site codon 148 itself, and four codons to the right of the A-site totalling 10 codons) was best correlated 149 with empirical data in their neural network framework. This window size is in line with 150 biological understanding of translational mechanisms, as the ribosome spans 151 approximately 10 codons along an mRNA strand during translation [23].…”

“…The data is then normalized on a per 93 gene basis, with a gene's raw RFP counts divided by the average RFP count in that 94 gene. These last four steps are inspired by [2] to help ensure data quality and 95 comparability across genes, and remove a total of 1,779 sequences (30%) from 96 consideration for the Tunney data and 1,173 sequences (19%) from the Weinberg data. 97 We also obtained 17 additional S. cerevisiae RFP data sets from 14 different studies 98 available from GWIPS-vis [18].…”

“…2 ribosomes at specific codon positions along mRNA sequences) has many biological 3 applications, such as enabling better understanding of co-translational protein folding 4 and aiding in gene design for heterologous expression. Ribosome footprinting (RFP, also 5 called ribosome profiling) is an experimental process often used to estimate ribosome 6March 26, 2020 1/16 tempo, i.e., the regional protein translation rate differences across a transcript [1,2].…”

“…March 26, 2020 1/16 tempo, i.e., the regional protein translation rate differences across a transcript [1,2].…”

XPRESSyourself: Enhancing, Standardizing, and Automating Ribosome Profiling Computational Analyses Yields Improved Insight into Data

“…RNA-Seq and ribosome-footprint coverage across these genes show that the significant changes in their TE are due to neither spurious, high-abundance fragments differentially present across libraries nor variance from an especially small number of mapped reads ( Figure S7). This is an important consideration as the commonly suggested use of the CircLigase enzyme in published ribosome profiling library preparation protocols, which circularizes template cDNA before sequencing, can bias certain molecules' incorporation into sequencing libraries based on read-end base content alone [83].…”