Decoding differential gene expression

Abstract

6Identifying the molecular mechanisms that control differential gene expression (DE) is a major goal of basic and disease 7 biology. Combining the strengths of systems biology and deep learning in a model called DEcode, we are able to predict 8 DE more accurately than traditional sequence-based methods, which do not utilize systems biology data. To determine 9 the biological origins of this accuracy, we identify the most predictive regulators and types of regulatory interactions in 10 DEcode, contrasting their roles across many human tissues. Diverse systems biology, ontological and disease-related 11 assessments all point to the predominant influence of post-translational RNA-binding factors on DE. Through the 12 combinatorial gene regulation that is captured in DEcode, it is even possible to predict relatively subtle person-to-person 13 variation in gene expression. We demonstrate the broad applicability of these clinically-relevant predictions by predicting 14 drivers of aging throughout the human lifespan, gene coexpression relationships on a genome-wide scale, and frequent 15 DE in diverse conditions. Researchers can freely access DEcode to utilize genomic big data in identifying influential 16 molecular mechanisms for any human expression data -www.differentialexpression.org. 17 18 While all human cells share DNA sequences, gene regulation differs among cell types and developmental stages, and in 19 response to environmental cues and stimuli. Accordingly, when gene expression is not properly regulated, cellular 20 homeostasis can be perturbed, often affecting cell function and leading to disease 1 . These distinctions between cell 21 states are observed as differential expression (DE) of gene transcripts. DE have been cataloged for tens of thousands of 22 gene expression datasets, in the context of distinctions between species, organs, and conditions. Despite the important 23These particularly challenging predictions are validated on a genome-wide scale, as we identify key drivers of 49 coexpression, and also drivers for phenotype-associated differential expression. These tests and applications indicate 50 DEcode can combine multiple recent data sources, to extract regulators for arbitrary human DE signatures. 51 3 Results 52Promoter and RNA features predict differential expression across human tissues 53The overarching goal of this study is to accurately predict gene expression as a function of molecular interactions. These 54 results should be tissue-specific, but also highlight major regulatory principles across tissues, and ideally have sufficient 55 accuracy to predict the relatively small expression changes observed between individual humans. To accomplish this, we 56 utilized deep convolutional neural networks in a system called DEcode that can predict inter-tissue variations and inter-57 person variations in gene expression levels from promoter and mRNA features (Figure 1). The promoter features 58 included: the genomic locations of binding sites of 762 TFs and the mRNA features encompassed the loca...

Sex-Specific Association of the X Chromosome With Cognitive Change and Tau Pathology in Aging and Alzheimer Disease

“…RNA-seq methods have been described in detail. 5 Of 13 822 coding genes detected genome-wide, 488 were from the X chromosome. Neuropathological examination and antemortem clinical and neuropsychological profiling were performed.…”

“…Neuropathological examination and antemortem clinical and neuropsychological profiling were performed. 5,6 Global cognitive function was derived for each individual from the annual neuropsychological evaluation, comprising 17 different tests that were collapsed to form rates of cognitive decline, controlling for age and years of education. 5 Cognitive decline was regressed against messenger RNA expression and covaried by extent of AD pathology (NFT and neuritic plaque scores); the association was defined as β.…”

“…5,6 Global cognitive function was derived for each individual from the annual neuropsychological evaluation, comprising 17 different tests that were collapsed to form rates of cognitive decline, controlling for age and years of education. 5 Cognitive decline was regressed against messenger RNA expression and covaried by extent of AD pathology (NFT and neuritic plaque scores); the association was defined as β. Brain NFT burden was regressed against messenger RNA expression, accounting for age and education; the association was defined as β.…”

“…3,4 With this in mind, we investigated an RNA sequencing (RNA-seq) data set from the well-established Religious Orders Study and Rush Memory and Aging Project joint cohorts to measure transcriptional levels of X gene expression in the dorsolateral prefrontal cortex, a cortical hub of multiple cognitive circuits, targeted by aging and AD. 5 Because X gene expression is imbalanced between the sexes, we performed separate analyses of women and men. In our main analysis, we examined whether X expression is associated with cognitive change during aging and AD, independent of AD pathology, in women and men.…”