Analysis of multiple types of omics data facilitates a comprehensive revelation of molecular-level complexity and interactions among genomic features. This knowledge promotes the development of new therapies for treating different genomic diseases. An integrative study of multiple types of genomic data instead of a single type of genomic data will be more informative in understanding the complicated molecular activities and their interactions. In this work, we integrated RNA-sequencing (RNA-seq), methylation, and DNA copy number variation data, downloaded from the TCGA public repository, of glioblastoma multiforme (GBM), reduced the dimension of these high-dimensional genomic data using an autoencoder, a deep learning-based method, and then used Cox-PH model to select the autoencoder-transformed features that have a significant contribution to patient survival. We utilized the significant set of autoencoder-transformed features to classify the survival subtypes using the integrated data. We built a classification model with a penalization technique, sparse group LASSO, and evaluated the approach using cross-validation. As a result, two survival subgroups, with overall different survival profiles and linking to various genomic features, are discovered for respective GBM patients. Finally, the results are interpreted biologically by differential expression analysis and pathway analysis.

DNA methylation and copy number may be associated with each other to some extent, in positive or negative ways. Whether differential methylation and copy number variation have combined effects on gene expression is largely unknown. We use a multivariate linear model to formulate the relationship among the three genomic measurements, gene expression, copy number, and methylation levels. We propose a method that combines a distance covariance measure and the group LASSO to analyze multiple types of genomic data collectively for the purpose of insightfully revealing how gene expression is potentially affected by both copy number variation and differential methylation levels in cellular process. Our approach is of two stages, the first is a variable screening process in which a variable selection method is utilized by employing the concept of joint distance covariance (JdCov) of random vectors, and the second is to implement a penalized regression approach, a group LASSO, on the screened data of much lower dimension. The two-stage approach is tested in extensive simulation studies and shown to be effective. The approach is then applied to the TCGA Melanoma data, which consists of gene expression, methylation and copy number measurements of more than 300 patients and relationship of some genes with methylation and copy number measurements were revealed for the involved subjects.

Array comparative genomic hybridization (aCGH) provides a high-resolution and high-throughput technique for screening of copy number variations (CNVs) within the entire genome. This technique, compared to the conventional CGH, significantly improves the identification of chromosomal abnormalities. However, due to the random noise inherited in the imaging and hybridization process, identifying statistically significant DNA copy number changes in aCGH data is challenging. We propose a novel approach that uses the mean and variance change point model (MVCM) to detect CNVs or breakpoints in aCGH data sets. We derive an approximate p-value for the test statistic and also give the estimate of the locus of the DNA copy number change. We carry out simulation studies to evaluate the accuracy of the estimate and the p-value formulation. These simulation results show that the approach is effective in identifying copy number changes. The approach is also tested on fibroblast cancer cell line data, breast tumor cell line data, and breast cancer cell line aCGH data sets that are publicly available. Changes that have not been identified by the circular binary segmentation (CBS) method but are biologically verified are detected by our approach on these cell lines with higher sensitivity and specificity than CBS.

BACKGROUND: Population structure is an important cause leading to inconsistent results in population-based association studies (PBAS) of human diseases. Various statistical methods have been proposed to reduce the negative impact of population structure on PBAS. Due to lack of structural information in real populations, it is difficult to evaluate the impact of population structure on PBAS in real populations. RESULTS: We developed a genetic simulation platform, HAPSIMU, based on real haplotype data from the HapMap ENCODE project. This platform can simulate heterogeneous populations with various known and controllable structures under the continuous migration model or the discrete model. Moreover, both qualitative and quantitative traits can be simulated using additive genetic model with various genetic parameters designated by users. CONCLUSION: HAPSIMU provides a common genetic simulation platform to evaluate the impact of population structure on PBAS, and compare the relative performance of various population structure identification and PBAS methods.

MOTIVATION: Periodic patterns in time series resulting from biological experiments are of great interest. The commonly used Fast Fourier Transform (FFT) algorithm is applicable only when data are evenly spaced and when no values are missing, which is not always the case in high-throughput measurements. The choice of statistic to evaluate the significance of the periodic patterns for unevenly spaced gene expression time series has not been well substantiated. METHODS: The Lomb-Scargle periodogram approach is used to search time series of gene expression to quantify the periodic behavior of every gene represented on the DNA array. The Lomb-Scargle periodogram analysis provides a direct method to treat missing values and unevenly spaced time points. We propose the combination of a Lomb-Scargle test statistic for periodicity and a multiple hypothesis testing procedure with controlled false discovery rate to detect significant periodic gene expression patterns. RESULTS: We analyzed the Plasmodium falciparum gene expression dataset. In the Quality Control Dataset of 5080 expression patterns, we found 4112 periodic probes. In addition, we identified 243 probes with periodic expression in the Complete Dataset, which could not be examined in the original study by the FFT analysis due to an excessive number of missing values. While most periodic genes had a period of 48 h, some had a period close to 24 h. Our approach should be applicable for detection and quantification of periodic patterns in any unevenly spaced gene expression time-series data.

BACKGROUND: One frequent application of microarray experiments is in the study of monitoring gene activities in a cell during cell cycle or cell division. A new challenge for analyzing the microarray experiments is to identify genes that are statistically significantly periodically expressed during the cell cycle. Such a challenge occurs due to the large number of genes that are simultaneously measured, a moderate to small number of measurements per gene taken at different time points, and high levels of non-normal random noises inherited in the data. RESULTS: Based on two statistical hypothesis testing methods for identifying periodic time series, a novel statistical inference approach, the C&G procedure, is proposed to effectively screen out statistically significantly periodically expressed genes. The approach is then applied to yeast and bacterial cell cycle gene expression data sets, as well as to human fibroblasts and human cancer cell line data sets, and significantly periodically expressed genes are successfully identified. CONCLUSION: The C&G procedure proposed is an effective method for identifying statistically significant periodic genes in microarray time series gene expression data.

On the subject of optimal subspaces for appearance-based object recognition, it is generally believed that algorithms based on LDA (linear discriminant analysis) are superior to those based on PCA (principal components analysis), provided that relatively large training data sets are available. In this paper, we show that while this is generally true for classification with the nearest-neighbor classifier, it is not always the case with a maximum-likelihood classifier. We support our claim by presenting both intuitively plausible arguments and actual results on a large data set of human chromosomes. Our conjecture is that perhaps only when the underlying object classes are linearly separable would LDA be truly superior to other known subspaces of equal dimensionality.