Volume 14, Issue 6

There is an increasing demand for exploration of the transcriptomes of multiple species with extraordinary traits such as the naked-mole rat (NMR). The NMR is remarkable because of its longevity and resistance to developing cancer. It is of scientific interest to understand the molecular mechanisms that impart these traits, and RNA-sequencing experiments with comparator species can correlate transcriptome dynamics with these phenotypes. Comparing transcriptome differences requires a homology mapping of each transcript in one species to transcript(s) within the other. Such mappings are necessary, especially if one species does not have well-annotated genome available. Current approaches for this type of analysis typically identify the best match for each transcript, but the best match analysis ignores the inherent risks of mismatch when there are multiple candidate transcripts with similar homology scores. We present a method that treats the set of homologs from a novel species as a cluster corresponding to a single gene in the reference species, and we compare the cluster-based approach to a conventional best-match analysis in both simulated data and a case study with NMR and mouse tissues. We demonstrate that the cluster-based approach has superior power to detect differential expression.

Genome-wide mapping of nucleosomes has revealed a great deal about the relationships between chromatin structure and control of gene expression. Recent next generation CHIP-chip and CHIP-Seq technologies have accelerated our understanding of basic principles of chromatin organization. These technologies have taught us that nucleosomes play a crucial role in gene regulation by allowing physical access to transcription factors. Recent methods and experimental advancements allow the determination of nucleosome positions for a given genome area. However, most of these methods estimate the number of nucleosomes either by an EM algorithm using a BIC criterion or an effective heuristic strategy. Here, we introduce a Bayesian method for identifying nucleosome positions. The proposed model is based on a Multinomial-Dirichlet classification and a hierarchical mixture distributions. The number and the positions of nucleosomes are estimated using a reversible jump Markov chain Monte Carlo simulation technique. We compare the performance of our method on simulated data and MNase-Seq data from Saccharomyces cerevisiae against PING and NOrMAL methods.

Splitting extended families into their component nuclear families to apply a genetic association method designed for nuclear families is a widespread practice in familial genetic studies. Dependence among genotypes and phenotypes of nuclear families from the same extended family arises because of genetic linkage of the tested marker with a risk variant or because of familial specificity of genetic effects due to gene-environment interaction. This raises concerns about the validity of inference conducted under the assumption of independence of the nuclear families. We indeed prove theoretically that, in a conditional logistic regression analysis applicable to disease cases and their genotyped parents, the naive model-based estimator of the variance of the coefficient estimates underestimates the true variance. However, simulations with realistic effect sizes of risk variants and variation of this effect from family to family reveal that the underestimation is negligible. The simulations also show the greater efficiency of the model-based variance estimator compared to a robust empirical estimator. Our recommendation is therefore, to use the model-based estimator of variance for inference on effects of genetic variants.

The rapidly developing sequencing technologies have led to improved disease risk prediction through identifying many novel genes. Many prediction methods have been proposed to use rich genomic information to predict binary disease outcomes. It is intuitive that these methods can be further improved by making efficient use of the rich information in measured quantitative traits that are correlated with binary outcomes. In this article, we propose a novel Empirical Bayes prediction model that uses information from both quantitative traits and binary disease status to improve risk prediction. Our method is built on a new statistic that better infers the gene effect on multiple traits, and it also enjoys the good theoretical properties. We then consider using sequencing data by combining information from multiple rare variants in individual genes to strengthen the signals of causal genetic effects. In simulation study, we find that our proposed Empirical Bayes approach is superior to other existing methods in terms of feature selection and risk prediction. We further evaluate the effectiveness of our proposed method through its application to the sequencing data provided by the Genetic Analysis Workshop 18.

The availability of large quantities of transcriptomic data in the form of RNA-seq count data has necessitated the development of methods to identify genes differentially expressed between experimental conditions. Many existing approaches apply a parametric model of gene expression and so place strong assumptions on the distribution of the data. Here we explore an alternate nonparametric approach that applies an empirical likelihood framework, allowing us to define likelihoods without specifying a parametric model of the data. We demonstrate the performance of our method when applied to gold standard datasets, and to existing experimental data. Our approach outperforms or closely matches performance of existing methods in the literature, and requires modest computational resources. An R package, EmpDiff implementing the methods described in the paper is available from: http://homepages.inf.ed.ac.uk/tthorne/software/packages/EmpDiff_0.99.tar.gz.