The ability to predict the impact of cis-regulatory sequence on gene expression would facilitate discovery in fundamental and applied biology. Here, we combine polysome profiling of a library of 280,000 randomized 5′ UTRs with deep learning to build a predictive model that relates human 5′ UTR sequence to translation. Together with a genetic algorithm, we use the model to engineer new 5′ UTRs that accurately direct specified levels of ribosome loading, providing the ability to tune sequences for optimal protein expression. We show that the same approach can be extended to chemically modified RNA, an important feature for applications in mRNA therapeutics and synthetic biology. We test 35,000 truncated human 5′ UTRs and 3,577 naturally occurring variants and show that the model predicts ribosome loading of these sequences. Finally, we provide evidence of 45 SNVs associated with human diseases that substantially change ribosome loading and thus may represent a molecular basis for disease. The sequence of the 5′ untranslated region (5′ UTR) is a primary determinant of translation efficiency 1,2. While many cis-regulatory elements within human 5′ UTRs have been characterized individually, the field still lacks a means to accurately predict protein expression from 5′ UTR sequence alone, limiting the ability to estimate the effects of genome-encoded variants and the ability to engineer 5′ UTRs for precise translation control. Massively parallel reporter assays (MPRAs)-methods that assess thousands to millions of sequence variants in a single experiment-coupled with machine learning have proven † Corresponding author. gseelig@uw.edu. Author contributions P.J.S and B.W. designed and performed experiments, performed data analysis and modeling, and wrote the manuscript. D.R. performed fluorescence validation experiments. V.P. and I.M. wrote the manuscript. D.R.M. helped design polysome profiling. G.S. designed experiments and wrote the manuscript.

RNA sequencing has been widely used as an essential tool to probe gene expression. While standard practices have been established to analyze RNA-seq data, it is still challenging to interpret and remove artifactual signals. Several biological and technical factors such as sex, age, batches, and sequencing technology have been found to bias these estimates. Probabilistic estimation of expression residuals (PEER), which infers broad variance components in gene expression measurements, has been used to account for some systematic effects, but it has remained challenging to interpret these PEER factors. Here we show that transcriptome diversity–a simple metric based on Shannon entropy–explains a large portion of variability in gene expression and is the strongest known factor encoded in PEER factors. We then show that transcriptome diversity has significant associations with multiple technical and biological variables across diverse organisms and datasets. In sum, transcriptome diversity provides a simple explanation for a major source of variation in both gene expression estimates and PEER covariates.

Although gene expression divergence has long been postulated to be the primary driver of human evolution, identifying the genes and genetic variants underlying uniquely human traits has proven to be quite challenging. Theory suggests that cell type-specific cis-regulatory variants may fuel evolutionary adaptation due to the specificity of their effects. These variants can precisely tune the expression of a single gene in a single cell type, avoiding the potentially deleterious consequences of trans-acting changes and non-cell type-specific changes that can impact many genes and cell types, respectively. It has recently become possible to quantify human-specific cis-acting regulatory divergence by measuring allele-specific expression in human-chimpanzee hybrid cells—the product of fusing induced pluripotent stem (iPS) cells of each species in vitro. However, these cis-regulatory changes have only been explored in a limited number of tissues and cell types. Here, we quantify human-chimpanzee cis-regulatory divergence in gene expression and chromatin accessibility across six cell types, enabling the identification of highly cell type-specific cis-regulatory changes. We find that cell type-specific genes and regulatory elements evolve faster than those shared across cell types, suggesting an important role for genes with cell type-specific expression in human evolution. Furthermore, we identify several instances of lineage-specific natural selection that may have played key roles in specific cell types, such as coordinated changes in the cis-regulation of dozens of genes involved in neuronal firing in motor neurons. Finally, using novel metrics and a machine learning model, we identify genetic variants that likely alter chromatin accessibility and transcription factor binding, leading to neuron-specific changes in the expression of the neurodevelopmentally important genes FABP7 and GAD1. Overall, our results demonstrate that integrative analysis of cis-regulatory divergence in chromatin accessibility and gene expression across cell types is a promising approach to identify the specific genes and genetic variants that make us human.

Although gene expression divergence has long been postulated to be the primary driver of human evolution, identifying the genes and genetic variants underlying uniquely human traits has proven to be quite challenging. Theory suggests that cell type-specificcis-regulatory variants may fuel evolutionary adaptation due to the specificity of their effects. These variants can precisely tune the expression of a single gene in a single cell type, avoiding the potentially deleterious consequences oftrans-acting changes and non-cell type-specific changes that can impact many genes and cell types, respectively. It has recently become possible to quantify human-specificcis-acting regulatory divergence by measuring allele-specific expression in human-chimpanzee hybrid cells—the product of fusing induced pluripotent stem (iPS) cells of each speciesin vitro. However, thesecis-regulatory changes have only been explored in a limited number of tissues and cell types. Here, we quantify human-chimpanzeecis-regulatory divergence in gene expression and chromatin accessibility across six cell types, enabling the identification of highly cell type-specificcis-regulatory changes. We find that cell type-specific genes and regulatory elements evolve faster than those shared across cell types, suggesting an important role for genes with cell type-specific expression in human evolution. Furthermore, we identify several instances of lineage-specific natural selection that may have played key roles in specific cell types, such as coordinated changes in thecis-regulation of dozens of genes involved in neuronal firing in motor neurons. Finally, using novel metrics and a machine learning model, we identify genetic variants that likely alter chromatin accessibility and transcription factor binding, leading to neuron-specific changes in the expression of the neurodevelopmentally important genesFABP7andGAD1. Overall, our results demonstrate that integrative analysis ofcis-regulatory divergence in chromatin accessibility and gene expression across cell types is a promising approach to identify the specific genes and genetic variants that make us human.

Although gene expression divergence has long been postulated to be the primary driver of human evolution, identifying the genes and genetic variants underlying uniquely human traits has proven to be quite challenging. Theory suggests that cell type-specific cis-regulatory variants may fuel evolutionary adaptation due to the specificity of their effects. These variants can precisely tune the expression of a single gene in a single cell type, avoiding the potentially deleterious consequences of trans-acting changes and non-cell type-specific changes that can impact many genes and cell types, respectively. It has recently become possible to quantify human-specific cis-acting regulatory divergence by measuring allele-specific expression in human-chimpanzee hybrid cells—the product of fusing induced pluripotent stem (iPS) cells of each species in vitro. However, these cis-regulatory changes have only been explored in a limited number of tissues and cell types. Here, we quantify human-chimpanzee cis-regulatory divergence in gene expression and chromatin accessibility across six cell types, enabling the identification of highly cell type-specific cis-regulatory changes. We find that cell type-specific genes and regulatory elements evolve faster than those shared across cell types, suggesting an important role for genes with cell type-specific expression in human evolution. Furthermore, we identify several instances of lineage-specific natural selection that may have played key roles in specific cell types, such as coordinated changes in the cis-regulation of dozens of genes involved in neuronal firing in motor neurons. Finally, using novel metrics and a machine learning model, we identify genetic variants that likely alter chromatin accessibility and transcription factor binding, leading to neuron-specific changes in the expression of the neurodevelopmentally important genes FABP7 and GAD1. Overall, our results demonstrate that integrative analysis of cis-regulatory divergence in chromatin accessibility and gene expression across cell types is a promising approach to identify the specific genes and genetic variants that make us human.