Predicting the translation efficiency of messenger RNA in mammalian cells

The mechanisms by which mRNA sequences specify translational control remain poorly understood in mammalian cells. Here we generate a transcriptome-wide atlas of translation efficiency (TE) measurements encompassing more than 140 human and mouse cell types from 3,819 ribosomal profiling datasets. We develop RiboNN, a state-of-the-art multitask deep convolutional neural network, and classic machine learning models to predict TEs in hundreds of cell types from sequence-encoded mRNA features. While most earlier models solely considered the 5' untranslated region (UTR) sequence, RiboNN integrates how the spatial positioning of low-level dinucleotide and trinucleotide features (that is, including codons) influences TE, capturing mechanistic principles such as how ribosomal processivity and tRNA abundance control translational output. RiboNN predicts the translational behavior of base-modified therapeutic RNA and explains evolutionary selection pressures in human 5' UTRs. Finally, it detects a common language governing mRNA regulatory control and highlights the interconnectedness of mRNA translation, stability and localization in mammalian organisms.