A deep recurrent neural network discovers complex biological rules to decipher RNA protein-coding potential.

The current deluge of newly identified RNA transcripts presents a singular opportunity for improved assessment of coding potential, a cornerstone of genome annotation, and for machine-driven discovery of biological knowledge. While traditional, feature-based methods for RNA classification are limited by current scientific knowledge, deep learning methods can independently discover complex biological rules in the data de novo. We trained a gated recurrent neural network (RNN) on human messenger RNA (mRNA) and long noncoding RNA (lncRNA) sequences. Our model, mRNA RNN (mRNN), surpasses state-of-the-art methods at predicting protein-coding potential despite being trained with less data and with no prior concept of what features define mRNAs. To understand what mRNN learned, we probed the network and uncovered several context-sensitive codons highly predictive of coding potential. Our results suggest that gated RNNs can learn complex and long-range patterns in full-length human transcripts, making them ideal for performing a wide range of difficult classification tasks and, most importantly, for harvesting new biological insights from the rising flood of sequencing data.

Transcription Factor CTIP1/ BCL11A Regulates Epidermal Differentiation and Lipid Metabolism During Skin Development.

The epidermal permeability barrier (EPB) prevents organisms from dehydration and infection. The transcriptional regulation of EPB development is poorly understood. We demonstrate here that transcription factor COUP-TF-interacting protein 1 (CTIP1/BCL11A; hereafter CTIP1) is highly expressed in the developing murine epidermis. Germline deletion of Ctip1 (Ctip1 -/-) results in EPB defects accompanied by compromised epidermal differentiation, drastic reduction in profilaggrin processing, reduced lamellar bodies in granular layers and significantly altered lipid composition. Transcriptional profiling of Ctip1 -/- embryonic skin identified altered expression of genes encoding lipid-metabolism enzymes, skin barrier-associated transcription factors and junctional proteins. CTIP1 was observed to interact with genomic elements within the regulatory region of the gene encoding the differentiation-associated gene, Fos-related antigen2 (Fosl2) and lipid-metabolism-related gene, Fatty acid elongase 4 (Elvol4), and the expression of both was altered in Ctip1 -/- mice. CTIP1 appears to play a role in EPB establishment of via direct or indirect regulation of a subset of genes encoding proteins involved in epidermal differentiation and lipid metabolism. These results identify potential, CTIP1-regulated avenues for treatment of skin disorders involving EBP defects.

MicroRNA and gene expression changes in unruptured human cerebral aneurysms.

OBJECTIVE The molecular mechanisms behind cerebral aneurysm formation and rupture remain poorly understood. In the past decade, microRNAs (miRNAs) have been shown to be key regulators in a host of biological processes. They are noncoding RNA molecules, approximately 21 nucleotides long, that posttranscriptionally inhibit mRNAs by attenuating protein translation and promoting mRNA degradation. The miRNA and mRNA interactions and expression levels in cerebral aneurysm tissue from human subjects were profiled. METHODS A prospective case-control study was performed on human subjects to characterize the differential expression of mRNA and miRNA in unruptured cerebral aneurysms in comparison with control tissue (healthy superficial temporal arteries [STA]). Ion Torrent was used for deep RNA sequencing. Affymetrix miRNA microarrays were used to analyze miRNA expression, whereas NanoString nCounter technology was used for validation of the identified targets. RESULTS Overall, 7 unruptured cerebral aneurysm and 10 STA specimens were collected. Several differentially expressed genes were identified in aneurysm tissue, with MMP-13 (fold change 7.21) and various collagen genes (COL1A1, COL5A1, COL5A2) being among the most upregulated. In addition, multiple miRNAs were significantly differentially expressed, with miR-21 (fold change 16.97) being the most upregulated, and miR-143-5p (fold change -11.14) being the most downregulated. From these, miR-21, miR-143, and miR-145 had several significantly anticorrelated target genes in the cohort that are associated with smooth muscle cell function, extracellular matrix remodeling, inflammation signaling, and lipid accumulation. All these processes are crucial to the pathophysiology of cerebral aneurysms. CONCLUSIONS This analysis identified differentially expressed genes and miRNAs in unruptured human cerebral aneurysms, suggesting the possibility of a role for miRNAs in aneurysm formation. Further investigation for their importance as therapeutic targets is needed.

How sequence defines structure: a crystallographic map of DNA structure and conformation.

The fundamental question of how sequence defines conformation is explicitly answered if the structures of all possible sequences of a macromolecule are determined. We present here a crystallographic screen of all permutations of the inverted repeat DNA sequence d(CCnnnN6N7N8GG), where N6, N7, and N8 are any of the four naturally occurring nucleotides. At this point, 63 of the 64 possible permutations have been crystallized from a defined set of solutions. When combined with previous work, we have assembled a data set of 37 single-crystal structures from 29 of the sequences in this motif, representing three structural classes of DNA (B-DNA, A-DNA, and four-stranded Holliday junctions). This data set includes a unique set of amphimorphic sequence, those that crystallize in two different conformations and serve to bridge the three structural phases. We have thus constructed a map of DNA structures that can be walked through in single nucleotide steps. Finally, the resulting data set allows us to dissect in detail the stabilization of and conformational variations within structural classes and identify significant conformational deviations within a particular structural class that result from sequence rather than crystal or crystallization effects.