Evaluating Metagenome Assembly on a Simple Defined Community with Many Strain Variants, bioRxiv, 2017-06-26
AbstractWe evaluate the performance of three metagenome assemblers, IDBA, MetaSPAdes, and MEGAHIT, on short-read sequencing of a defined “mock” community containing 64 genomes (Shakya et al. (2013)). We update the reference metagenome for this mock community and detect several additional genomes in the read data set. We show that strain confusion results in significant loss in assembly of reference genomes that are otherwise completely present in the read data set. In agreement with previous studies, we find that MEGAHIT performs best computationally; we also show that MEGAHIT tends to recover larger portions of the strain variants than the other assemblers.
biorxiv bioinformatics 100-200-users 2017Biological classification with RNA-Seq data Can alternative splicing enhance machine learning classifier?, bioRxiv, 2017-06-19
AbstractThe extent to which the genes are expressed in the cell can be simplistically defined as a function of one or more factors of the environment, lifestyle, and genetics. RNA sequencing (RNA-Seq) is becoming a prevalent approach to quantify gene expression, and is expected to gain better insights to a number of biological and biomedical questions, compared to the DNA microarrays. Most importantly, RNA-Seq allows to quantify expression at the gene and alternative splicing isoform levels. However, leveraging the RNA-Seq data requires development of new data mining and analytics methods. Supervised machine learning methods are commonly used approaches for biological data analysis, and have recently gained attention for their applications to the RNA-Seq data.In this work, we assess the utility of supervised learning methods trained on RNA-Seq data for a diverse range of biological classification tasks. We hypothesize that the isoform-level expression data is more informative for biological classification tasks than the gene-level expression data. Our large-scale assessment is done through utilizing multiple datasets, organisms, lab groups, and RNA-Seq analysis pipelines. Overall, we performed and assessed 61 biological classification problems that leverage three independent RNA-Seq datasets and include over 2,000 samples that come from multiple organisms, lab groups, and RNA-Seq analyses. These 61 problems include predictions of the tissue type, sex, or age of the sample, healthy or cancerous phenotypes and, the pathological tumor stage for the samples from the cancerous tissue. For each classification problem, the performance of three normalization techniques and six machine learning classifiers was explored. We find that for every single classification problem, the isoform-based classifiers outperform or are comparable with gene expression based methods. The top-performing supervised learning techniques reached a near perfect classification accuracy, demonstrating the utility of supervised learning for RNA-Seq based data analysis.
biorxiv bioinformatics 0-100-users 2017Graphtyper Population-scale genotyping using pangenome graphs, bioRxiv, 2017-06-10
AbstractA fundamental requisite for genetic studies is an accurate determination of sequence variation. While human genome sequence diversity is increasingly well characterized, there is a need for efficient ways to utilize this knowledge in sequence analysis. Here we present Graphtyper, a publicly available novel algorithm and software for discovering and genotyping sequence variants. Graphtyper realigns short-read sequence data to a pangenome, a variation-aware graph structure that encodes sequence variation within a population by representing possible haplotypes as graph paths. Our results show that Graphtyper is fast, highly scalable, and provides sensitive and accurate genotype calls. Graphtyper genotyped 89.4 million sequence variants in whole-genomes of 28,075 Icelanders using less than 100,000 CPU days, including detailed genotyping of six human leukocyte antigen (HLA) genes. We show that Graphtyper is a valuable tool in characterizing sequence variation in population-scale sequencing studies.
biorxiv bioinformatics 0-100-users 2017Integrating long-range connectivity information into de Bruijn graphs, bioRxiv, 2017-06-09
AbstractMotivationThe de Bruijn graph is a simple and efficient data structure that is used in many areas of sequence analysis including genome assembly, read error correction and variant calling. The data structure has a single parameter k, is straightforward to implement and is tractable for large genomes with high sequencing depth. It also enables representation of multiple samples simultaneously to facilitate comparison. However, unlike the string graph, a de Bruijn graph does not retain long range information that is inherent in the read data. For this reason, applications that rely on de Bruijn graphs can produce sub-optimal results given their input.ResultsWe present a novel assembly graph data structure the Linked de Bruijn Graph (LdBG). Constructed by adding annotations on top of a de Bruijn graph, it stores long range connectivity information through the graph. We show that with error-free data it is possible to losslessly store and recover sequence from a Linked de Bruijn graph. With assembly simulations we demonstrate that the LdBG data structure outperforms both the de Bruijn graph and the String Graph Assembler (SGA). Finally we apply the LdBG to Klebsiella pneumoniae short read data to make large (12 kbp) variant calls, which we validate using PacBio sequencing data, and to characterise the genomic context of drug-resistance genes.AvailabilityLinked de Bruijn Graphs and associated algorithms are implemented as part of McCortex, available under the MIT license at <jatsext-link xmlnsxlink=httpwww.w3.org1999xlink ext-link-type=uri xlinkhref=httphttpsgithub.commcveanmccortex>httpsgithub.commcveanmccortex<jatsext-link>.Contactturner.isaac@gmail.com.
biorxiv bioinformatics 0-100-users 2017Improving the value of public RNA-seq expression data by phenotype prediction, bioRxiv, 2017-06-04
Abstract<jatssec id=sa1>BackgroundPublicly available genomic data are a valuable resource for studying normal human variation and disease, but these data are often not well labeled or annotated. The lack of phenotype information for public genomic data severely limits their utility for addressing targeted biological questions.<jatssec id=sa2>ResultsWe develop an in silico phenotyping approach for predicting critical missing annotation directly from genomic measurements using, well-annotated genomic and phenotypic data produced by consortia like TCGA and GTEx as training data. We apply in silico phenotyping to a set of 70,000 RNA-seq samples we recently processed on a common pipeline as part of the recount2 project (<jatsext-link xmlnsxlink=httpwww.w3.org1999xlink ext-link-type=uri xlinkhref=httpsjhubiostatistics.shinyapps.iorecount>httpsjhubiostatistics.shinyapps.iorecount<jatsext-link>). We use gene expression data to build and evaluate predictors for both biological phenotypes (sex, tissue, sample source) and experimental conditions (sequencing strategy). We demonstrate how these predictions can be used to study cross-sample properties of public genomic data, select genomic projects with specific characteristics, and perform downstream analyses using predicted phenotypes. The methods to perform phenotype prediction are available in the phenopredict R package (<jatsext-link xmlnsxlink=httpwww.w3.org1999xlink ext-link-type=uri xlinkhref=httpsgithub.comleekgroupphenopredict>httpsgithub.comleekgroupphenopredict<jatsext-link>) and the predictions for recount2 are available from the recount R package (<jatsext-link xmlnsxlink=httpwww.w3.org1999xlink ext-link-type=uri xlinkhref=httpsbioconductor.orgpackagesreleasebiochtmlrecount.html>httpsbioconductor.orgpackagesreleasebiochtmlrecount.html<jatsext-link>)<jatssec id=sa3>ConclusionHaving leveraging massive public data sets to generate a well-phenotyped set of expression data for more than 70,000 human samples, expression data is available for use on a scale that was not previously feasible.
biorxiv bioinformatics 100-200-users 2017SCENIC Single-cell regulatory network inference and clustering, bioRxiv, 2017-06-01
AbstractSingle-cell RNA-seq allows building cell atlases of any given tissue and infer the dynamics of cellular state transitions during developmental or disease trajectories. Both the maintenance and transitions of cell states are encoded by regulatory programs in the genome sequence. However, this regulatory code has not yet been exploited to guide the identification of cellular states from single-cell RNA-seq data. Here we describe a computational resource, called SCENIC (Single Cell rEgulatory Network Inference and Clustering), for the simultaneous reconstruction of gene regulatory networks (GRNs) and the identification of stable cell states, using single-cell RNA-seq data. SCENIC outperforms existing approaches at the level of cell clustering and transcription factor identification. Importantly, we show that cell state identification based on GRNs is robust towards batch-effects and technical-biases. We applied SCENIC to a compendium of single-cell data from the mouse and human brain and demonstrate that the proper combinations of transcription factors, target genes, enhancers, and cell types can be identified. Moreover, we used SCENIC to map the cell state landscape in melanoma and identified a gene regulatory network underlying a proliferative melanoma state driven by MITF and STAT and a contrasting network controlling an invasive state governed by NFATC2 and NFIB. We further validated these predictions by showing that two transcription factors are predominantly expressed in early metastatic sentinel lymph nodes. In summary, SCENIC is the first method to analyze scRNA-seq data using a network-centric, rather than cell-centric approach. SCENIC is generic, easy to use, and flexible, and allows for the simultaneous tracing of genomic regulatory programs and the mapping of cellular identities emerging from these programs. Availability SCENIC is available as an R workflow based on three new RBioconductor packages GENIE3, RcisTarget and AUCell. As scalable alternative to GENIE3, we also provide GRNboost, paving the way towards the network analysis across millions of single cells.
biorxiv bioinformatics 0-100-users 2017