Explore the vast range of titles available.

How noncoding DNA determines gene expression in different cell types is a major unsolved problem, and critical downstream applications in human genetics depend on improved solutions. Here, we report substantially improved gene expression prediction accuracy from DNA sequences through the use of a deep learning architecture, called Enformer, that is able to integrate information from long-range interactions (up to 100 kb away) in the genome. This improvement yielded more accurate variant effect predictions on gene expression for both natural genetic variants and saturation mutagenesis measured by massively parallel reporter assays. Furthermore, Enformer learned to predict enhancer–promoter interactions directly from the DNA sequence competitively with methods that take direct experimental data as input. We expect that these advances will enable more effective fine-mapping of human disease associations and provide a framework to interpret cis-regulatory evolution.By using a new deep learning architecture, Enformer leverages long-range information to improve prediction of gene expression on the basis of DNA sequence.

Long non-protein-coding RNAs (lncRNAs) are proposed to be the largest transcript class in the mouse and human transcriptomes. Two important questions are whether all lncRNAs are functional and how they could exert a function. Several lncRNAs have been shown to function through their product, but this is not the only possible mode of action. In this review we focus on a role for the process of lncRNA transcription, independent of the lncRNA product, in regulating protein-coding-gene activity in cis . We discuss examples where lncRNA transcription leads to gene silencing or activation, and describe strategies to determine if the lncRNA product or its transcription causes the regulatory effect.

The androgen receptor (AR) plays a critical role in the development of the normal prostate as well as prostate cancer. Using an integrative transcriptomic analysis of prostate cancer cell lines and tissues, we identified ARLNC1 (AR-regulated long noncoding RNA 1) as an important long noncoding RNA that is strongly associated with AR signaling in prostate cancer progression. Not only was ARLNC1 induced by the AR protein, but ARLNC1 stabilized the AR transcript via RNA–RNA interaction. ARLNC1 knockdown suppressed AR expression, global AR signaling and prostate cancer growth in vitro and in vivo. Taken together, these data support a role for ARLNC1 in maintaining a positive feedback loop that potentiates AR signaling during prostate cancer progression and identify ARLNC1 as a novel therapeutic target. ARLNC1 is a newly discovered lncRNA that is induced by androgen receptor (AR) and maintains AR signaling by stabilizing the AR transcript. Knockdown of ARLNC1 suppresses AR expression, AR signaling and prostate cancer growth in vitro and in vivo.

The X-ray crystal structure and biochemical analysis of a triple helix formed between the expression and nuclear retention element (ENE) and the 3′ poly(A) tail of the human long noncoding RNA MALAT-1 reveals the basis of its stability and how it confers resistance to degradation. Metastasis-associated lung adenocarcinoma transcript 1 (MALAT1) is a highly abundant nuclear long noncoding RNA that promotes malignancy. A 3′-stem-loop structure is predicted to confer stability by engaging a downstream A-rich tract in a triple helix, similar to the expression and nuclear retention element (ENE) from the KSHV polyadenylated nuclear RNA. The 3.1-Å-resolution crystal structure of the human MALAT1 ENE and A-rich tract reveals a bipartite triple helix containing stacks of five and four U•A-U triples separated by a C + •G-C triplet and C-G doublet, extended by two A-minor interactions. In vivo decay assays indicate that this blunt-ended triple helix, with the 3′ nucleotide in a U•A-U triple, inhibits rapid nuclear RNA decay. Interruption of the triple helix by the C-G doublet induces a 'helical reset' that explains why triple-helical stacks longer than six do not occur in nature.

DeepSEA, a deep-learning algorithm trained on large-scale chromatin-profiling data, predicts chromatin effects from sequence alone, has single-nucleotide sensitivity and can predict effects of noncoding variants. Identifying functional effects of noncoding variants is a major challenge in human genetics. To predict the noncoding-variant effects de novo from sequence, we developed a deep learning–based algorithmic framework, DeepSEA ( http://deepsea.princeton.edu/ ), that directly learns a regulatory sequence code from large-scale chromatin-profiling data, enabling prediction of chromatin effects of sequence alterations with single-nucleotide sensitivity. We further used this capability to improve prioritization of functional variants including expression quantitative trait loci (eQTLs) and disease-associated variants.

Noncoding repeat expansions cause various neuromuscular diseases, including myotonic dystrophies, fragile X tremor/ataxia syndrome, some spinocerebellar ataxias, amyotrophic lateral sclerosis and benign adult familial myoclonic epilepsies. Inspired by the striking similarities in the clinical and neuroimaging findings between neuronal intranuclear inclusion disease (NIID) and fragile X tremor/ataxia syndrome caused by noncoding CGG repeat expansions in FMR1 , we directly searched for repeat expansion mutations and identified noncoding CGG repeat expansions in NBPF19 ( NOTCH2NLC ) as the causative mutations for NIID. Further prompted by the similarities in the clinical and neuroimaging findings with NIID, we identified similar noncoding CGG repeat expansions in two other diseases: oculopharyngeal myopathy with leukoencephalopathy and oculopharyngodistal myopathy, in LOC642361 / NUTM2B-AS1 and LRP12 , respectively. These findings expand our knowledge of the clinical spectra of diseases caused by expansions of the same repeat motif, and further highlight how directly searching for expanded repeats can help identify mutations underlying diseases. Whole-genome sequencing identifies noncoding CGG repeat expansions in neuronal intranuclear inclusion disease, oculopharyngodistal myopathy and oculopharyngeal myopathy with leukoencephalopathy, three disorders with overlapping clinical features and neuroimaging findings.

Coding variants represent many of the strongest associations between genotype and phenotype; however, they exhibit inter-individual differences in effect, termed ‘variable penetrance’. Here, we study how cis-regulatory variation modifies the penetrance of coding variants. Using functional genomic and genetic data from the Genotype-Tissue Expression Project (GTEx), we observed that in the general population, purifying selection has depleted haplotype combinations predicted to increase pathogenic coding variant penetrance. Conversely, in cancer and autism patients, we observed an enrichment of penetrance increasing haplotype configurations for pathogenic variants in disease-implicated genes, providing evidence that regulatory haplotype configuration of coding variants affects disease risk. Finally, we experimentally validated this model by editing a Mendelian single-nucleotide polymorphism (SNP) using CRISPR/Cas9 on distinct expression haplotypes with the transcriptome as a phenotypic readout. Our results demonstrate that joint regulatory and coding variant effects are an important part of the genetic architecture of human traits and contribute to modified penetrance of disease-causing variants. Analysis of GTEx, cancer and autism data sets shows that cis-regulatory variation can modify the penetrance of coding variants. Deleterious coding variants on regulatory haplotypes resulting in high expression are enriched in disease cohorts and selected against in general populations.

The genome-wide annotation of variants (GWAVA) software predicts whether noncoding variants are likely to be functional using a classifier trained on a range of genomic and epigenomic annotations. Identifying functionally relevant variants against the background of ubiquitous genetic variation is a major challenge in human genetics. For variants in protein-coding regions, our understanding of the genetic code and splicing allows us to identify likely candidates, but interpreting variants outside genic regions is more difficult. Here we present genome-wide annotation of variants (GWAVA), a tool that supports prioritization of noncoding variants by integrating various genomic and epigenomic annotations.