BACKGROUND: Parent-of-origin allele-specific gene expression (ASE) can be detected in interspecies hybrids by virtue of RNA sequence variants between the parental haplotypes. ASE is detectable by differential expression analysis (DEA) applied to the counts of RNA-seq read pairs aligned to parental references, but aligners do not always choose the correct parental reference. RESULTS: We used public data for species that are known to hybridize. We measured our ability to assign RNA-seq read pairs to their proper transcriptome or genome references. We tested software packages that assign each read pair to a reference position and found that they often favored the incorrect species reference. To address this problem, we introduce a post process that extracts alignment features and trains a random forest classifier to choose the better alignment. On each simulated hybrid dataset tested, our machine-learning post-processor achieved higher accuracy than the aligner by itself at choosing the correct parent-of-origin per RNA-seq read pair. CONCLUSIONS: For the parent-of-origin classification of RNA-seq, machine learning can improve the accuracy of alignment-based methods. This approach could be useful for enhancing ASE detection in interspecies hybrids, though RNA-seq from real hybrids may present challenges not captured by our simulations. We believe this is the first application of machine learning to this problem domain.

In this work, we studied the question of whether the two ends of long non-coding ribonucleic acids (lncRNAs) (i.e., the 5′ end and 3′end) carry similar information about subcellular localization of lncRNAs. We considered this problem from three viewpoints using machine learning models: (1) consideration of the classification performance of the machine learning models using features from defined regions (or segments) along the sequence, (2) correlation-based analysis using models built on regions/segments along the LncRNA sequence, and (3) analysis of the relative positions of predicted lncRNA localization motifs along the LncRNA sequence. Our results and observations suggest that the 5′ region of the lncRNA sequences (the prefixes) tend to carry more localization signals when compared with the 3′ region (the suffixes) of the sequences. These could have implications on how we use machine learning models for improved analysis of lncRNA subcellular localization.

Distinguishing long non-coding RNA from protein-coding RNA is important to molecular and cellular biology. The problem can be addressed with machine learning in general and with artificial neural networks in particular. We explore the effects of various network design choices on the accuracy of human LncRNA identification. Perceptron-based neural network models were found to be almost as accurate as more complex recurrent neural networks, and K-mer representations of the data seemed to assist both. Size selection of training data affected results. These explorations could assist in neural network design for RNA analysis.

Summary Araport is an open‐source, online community resource for research on the Arabidopsis thaliana genome and related data. Araport is developed through a partnership between J. Craig Venter Institute, the Texas Advanced Computing Center at The University of Texas at Austin, and The University of Cambridge. Part of the open architecture of Araport is the Science Applications Workspace. Taking an ‘app store’ approach, users can choose applications developed both by the Araport team and community developers to create a customized environment for their work. Araport also provides tooling and support for developing applications for Araport, including an application generator, a rapid development and testing tool, and a straightforward deployment path for publishing applications into the Araport workspace. Copyright © 2015 John Wiley & Sons, Ltd.

BACKGROUND: Finishing is the process of improving the quality and utility of draft genome sequences generated by shotgun sequencing and computational assembly. Finishing can involve targeted sequencing. Finishing reads may be incorporated by manual or automated means. One automated method uses targeted addition by local re-assembly of gap regions. An obvious alternative uses de novo assembly of all the reads. RESULTS: A procedure called the bounding read algorithm was developed for assembly of shotgun reads plus finishing reads and their constraints, targeting repeat regions. The algorithm was implemented within the Celera Assembler software and its pyrosequencing-specific variant, CABOG. The implementation was tested on Sanger and pyrosequencing data from six genomes. The bounding read assemblies were compared to assemblies from two other methods on the same data. The algorithm generates improved assemblies of repeat regions, closing and tiling some gaps while degrading none. CONCLUSIONS: The algorithm is useful for small-genome automated finishing projects. Our implementation is available as open-source from http://wgs-assembler.sourceforge.net under the GNU Public License.

MOTIVATION: We present an algorithm to identify allelic variation given a Whole Genome Shotgun (WGS) assembly of haploid sequences, and to produce a set of haploid consensus sequences rather than a single consensus sequence. Existing WGS assemblers take a column-by-column approach to consensus generation, and produce a single consensus sequence which can be inconsistent with the underlying haploid alleles, and inconsistent with any of the aligned sequence reads. Our new algorithm uses a dynamic windowing approach. It detects alleles by simultaneously processing the portions of aligned reads spanning a region of sequence variation, assigns reads to their respective alleles, phases adjacent variant alleles and generates a consensus sequence corresponding to each confirmed allele. This algorithm was used to produce the first diploid genome sequence of an individual human. It can also be applied to assemblies of multiple diploid individuals and hybrid assemblies of multiple haploid organisms. RESULTS: Being applied to the individual human genome assembly, the new algorithm detects exactly two confirmed alleles and reports two consensus sequences in 98.98% of the total number 2,033311 detected regions of sequence variation. In 33,269 out of 460,373 detected regions of size >1 bp, it fixes the constructed errors of a mosaic haploid representation of a diploid locus as produced by the original Celera Assembler consensus algorithm. Using an optimized procedure calibrated against 1 506 344 known SNPs, it detects 438 814 new heterozygous SNPs with false positive rate 12%. AVAILABILITY: The open source code is available at: http://wgs-assembler.cvs.sourceforge.net/wgs-assembler/

MOTIVATION: DNA sequence reads from Sanger and pyrosequencing platforms differ in cost, accuracy, typical coverage, average read length and the variety of available paired-end protocols. Both read types can complement one another in a 'hybrid' approach to whole-genome shotgun sequencing projects, but assembly software must be modified to accommodate their different characteristics. This is true even of pyrosequencing mated and unmated read combinations. Without special modifications, assemblers tuned for homogeneous sequence data may perform poorly on hybrid data. RESULTS: Celera Assembler was modified for combinations of ABI 3730 and 454 FLX reads. The revised pipeline called CABOG (Celera Assembler with the Best Overlap Graph) is robust to homopolymer run length uncertainty, high read coverage and heterogeneous read lengths. In tests on four genomes, it generated the longest contigs among all assemblers tested. It exploited the mate constraints provided by paired-end reads from either platform to build larger contigs and scaffolds, which were validated by comparison to a finished reference sequence. A low rate of contig mis-assembly was detected in some CABOG assemblies, but this was reduced in the presence of sufficient mate pair data. AVAILABILITY: The software is freely available as open-source from http://wgs-assembler.sf.net under the GNU Public License.