We conducted a systematic assessment of computational models-CellDMC, TCA, HIRE, TOAST, and CeDAR-for detecting cell-type-specific differential methylation CpGs in bulk methylation data profiled using the Illumina DNA Methylation BeadArrays. This assessment was performed through simulations and case studies involving two epigenome-wide association studies (EWAS) on rheumatoid arthritis and major depressive disorder. Our evaluation provided insights into the strengths and limitations of each model. The results revealed that the models varied in performance across different metrics, sample sizes, and computational efficiency. Additionally, we proposed integrating the results from these models using the minimum p-value ($minpv$) and average p-value ($avepv$) approaches. Our findings demonstrated that these aggregation methods significantly improved performance in identifying cell-type-specific differential methylation CpGs.

We performed systematic assessment of computational deconvolution methods that play an important role in the estimation of cell type proportions from bulk methylation data. The proposed framework methylDeConv (available as an R package) integrates several deconvolution methods for methylation profiles (Illumina HumanMethylation450 and MethylationEPIC arrays) and offers different cell-type-specific CpG selection to construct the extended reference library which incorporates the main immune cell subsets, epithelial cells and cell-free DNAs. We compared the performance of different deconvolution algorithms via simulations and benchmark datasets and further investigated the associations of the estimated cell type proportions to cancer therapy in breast cancer and subtypes in melanoma methylation case studies. Our results indicated that the deconvolution based on the extended reference library is critical to obtain accurate estimates of cell proportions in non-blood tissues.

MOTIVATION: A gradient boosting decision tree (GBDT) is a powerful ensemble machine-learning method that has the potential to accelerate biomarker discovery from high-dimensional molecular data. Recent algorithmic advances, such as extreme gradient boosting (XGB) and light gradient boosting (LGB), have rendered the GBDT training more efficient, scalable and accurate. However, these modern techniques have not yet been widely adopted in discovering biomarkers for censored survival outcomes, which are key clinical outcomes or endpoints in cancer studies. RESULTS: In this paper, we present a new R package 'Xsurv' as an integrated solution that applies two modern GBDT training frameworks namely, XGB and LGB, for the modeling of right-censored survival outcomes. Based on our simulations, we benchmark the new approaches against traditional methods including the stepwise Cox regression model and the original gradient boosting function implemented in the package 'gbm'. We also demonstrate the application of Xsurv in analyzing a melanoma methylation dataset. Together, these results suggest that Xsurv is a useful and computationally viable tool for screening a large number of prognostic candidate biomarkers, which may facilitate future translational and clinical research. AVAILABILITY AND IMPLEMENTATION: 'Xsurv' is freely available as an R package at: https://github.com/topycyao/Xsurv. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

The Cox proportional hazard model is one of the most widely used methods in modeling time-to-event data in the health sciences. Due to the simplicity of the Cox partial likelihood function, many machine learning algorithms use it for survival data. However, due to the nature of censored data, the optimization problem becomes intractable when more complicated regularization is employed, which is necessary when dealing with high dimensional omic data. In this paper, we show that a convex conjugate function of the Cox loss function based on Fenchel duality exists, and provide an alternative framework to optimization based on the primal form. Furthermore, the dual form suggests an efficient algorithm for solving the kernel learning problem with censored survival outcomes. We illustrate performance and properties of the derived duality form of Cox partial likelihood loss in multiple kernel learning problems with simulated and the Skin Cutaneous Melanoma TCGA datasets.

BACKGROUND: High-throughput sequencing experiments followed by differential expression analysis is a widely used approach for detecting genomic biomarkers. A fundamental step in differential expression analysis is to model the association between gene counts and covariates of interest. Existing models assume linear effect of covariates, which is restrictive and may not be sufficient for certain phenotypes. RESULTS: We introduce NBAMSeq, a flexible statistical model based on the generalized additive model and allows for information sharing across genes in variance estimation. Specifically, we model the logarithm of mean gene counts as sums of smooth functions with the smoothing parameters and coefficients estimated simultaneously within a nested iterative method. The variance is estimated by the Bayesian shrinkage approach to fully exploit the information across all genes. CONCLUSIONS: Based on extensive simulations and case studies of RNA-Seq data, we show that NBAMSeq offers improved performance in detecting nonlinear effect and maintains equivalent performance in detecting linear effect compared to existing methods. The vignette and source code of NBAMSeq are available at http://bioconductor.org/packages/release/bioc/html/NBAMSeq.html.

MOTIVATION: An important downstream analysis following differential expression from RNA sequencing (RNA-Seq) or DNA methylation analysis is the gene set testing to relate significant genes or CpGs to known biological properties. However, the traditional gene set testing approaches result in biased P-values due to the difference in gene length. Existing methods accounting for length bias were primarily developed for RNA-Seq data. For DNA methylation data profiled using the Illumina arrays, separate methods adjusting for the number of CpGs instead of gene length are necessary. RESULTS: We developed methylGSA, a Bioconductor package for gene set testing in DNA methylation data. Our accompanying Shiny app provides an interactive way of accessing functions and visualizing the results in methylGSA package. AVAILABILITY AND IMPLEMENTATION: methylGSA is available at Bioconductor repository: https://bioconductor.org/packages/methylGSA and Shiny app is available at: http://www.ams.sunysb.edu/%7epfkuan/softwares.html#methylGSA. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Advances in medical technology have allowed for customized prognosis, diagnosis, and treatment regimens that utilize multiple heterogeneous data sources. Multiple kernel learning (MKL) is well suited for the integration of multiple high throughput data sources. MKL remains to be under-utilized by genomic researchers partly due to the lack of unified guidelines for its use, and benchmark genomic datasets. RESULTS: We provide three implementations of MKL in R. These methods are applied to simulated data to illustrate that MKL can select appropriate models. We also apply MKL to combine clinical information with miRNA gene expression data of ovarian cancer study into a single analysis. Lastly, we show that MKL can identify gene sets that are known to play a role in the prognostic prediction of 15 cancer types using gene expression data from The Cancer Genome Atlas, as well as, identify new gene sets for the future research. CONCLUSION: Multiple kernel learning coupled with modern optimization techniques provides a promising learning tool for building predictive models based on multi-source genomic data. MKL also provides an automated scheme for kernel prioritization and parameter tuning. The methods used in the paper are implemented as an R package called RMKL package, which is freely available for download through CRAN at https://CRAN.R-project.org/package=RMKL .

BACKGROUND: CRISPR is a versatile gene editing tool which has revolutionized genetic research in the past few years. Optimizing sgRNA design to improve the efficiency of target/DNA cleavage is critical to ensure the success of CRISPR screens. RESULTS: By borrowing knowledge from oligonucleotide design and nucleosome occupancy models, we systematically evaluated candidate features computed from a number of nucleic acid, thermodynamic and secondary structure models on real CRISPR datasets. Our results showed that taking into account position-dependent dinucleotide features improved the design of effective sgRNAs with area under the receiver operating characteristic curve (AUC) >0.8, and the inclusion of additional features offered marginal improvement (∼2% increase in AUC). CONCLUSION: Using a machine-learning approach, we proposed an accurate prediction model for sgRNA design efficiency. An R package predictSGRNA implementing the predictive model is available at http://www.ams.sunysb.edu/~pfkuan/softwares.html#predictsgrna .

Chromatin immunoprecipitation followed by high-throughput sequencing (ChIP-seq) is rapidly replacing chromatin immunoprecipitation combined with genome-wide tiling array analysis (ChIP-chip) as the preferred approach for mapping transcription-factor binding sites and chromatin modifications. The state of the art for analyzing ChIP-seq data relies on using only reads that map uniquely to a relevant reference genome (uni-reads). This can lead to the omission of up to 30% of alignable reads. We describe a general approach for utilizing reads that map to multiple locations on the reference genome (multi-reads). Our approach is based on allocating multi-reads as fractional counts using a weighted alignment scheme. Using human STAT1 and mouse GATA1 ChIP-seq datasets, we illustrate that incorporation of multi-reads significantly increases sequencing depths, leads to detection of novel peaks that are not otherwise identifiable with uni-reads, and improves detection of peaks in mappable regions. We investigate various genome-wide characteristics of peaks detected only by utilization of multi-reads via computational experiments. Overall, peaks from multi-read analysis have similar characteristics to peaks that are identified by uni-reads except that the majority of them reside in segmental duplications. We further validate a number of GATA1 multi-read only peaks by independent quantitative real-time ChIP analysis and identify novel target genes of GATA1. These computational and experimental results establish that multi-reads can be of critical importance for studying transcription factor binding in highly repetitive regions of genomes with ChIP-seq experiments.

MOTIVATION: The Illumina BeadArray is a popular platform for profiling DNA methylation, an important epigenetic event associated with gene silencing and chromosomal instability. However, current approaches rely on an arbitrary detection P-value cutoff for excluding probes and samples from subsequent analysis as a quality control step, which results in missing observations and information loss. It is desirable to have an approach that incorporates the whole data, but accounts for the different quality of individual observations. RESULTS: We first investigate and propose a statistical framework for removing the source of biases in Illumina Methylation BeadArray based on several positive control samples. We then introduce a weighted model-based clustering called LumiWCluster for Illumina BeadArray that weights each observation according to the detection P-values systematically and avoids discarding subsets of the data. LumiWCluster allows for discovery of distinct methylation patterns and automatic selection of informative CpG loci. We demonstrate the advantages of LumiWCluster on two publicly available Illumina GoldenGate Methylation datasets (ovarian cancer and hepatocellular carcinoma). AVAILABILITY: R package LumiWCluster can be downloaded from http://www.unc.edu/∼pfkuan/LumiWCluster.

BACKGROUND: Chromatin immunoprecipitation combined with DNA microarrays (ChIP-chip) is an assay used for investigating DNA-protein-binding or post-translational chromatin/histone modifications. As with all high-throughput technologies, it requires thorough bioinformatic processing of the data for which there is no standard yet. The primary goal is to reliably identify and localize genomic regions that bind a specific protein. Further investigation compares binding profiles of functionally related proteins, or binding profiles of the same proteins in different genetic backgrounds or experimental conditions. Ultimately, the goal is to gain a mechanistic understanding of the effects of DNA binding events on gene expression. RESULTS: We present a free, open-source R/Bioconductor package Starr that facilitates comparative analysis of ChIP-chip data across experiments and across different microarray platforms. The package provides functions for data import, quality assessment, data visualization and exploration. Starr includes high-level analysis tools such as the alignment of ChIP signals along annotated features, correlation analysis of ChIP signals with complementary genomic data, peak-finding and comparative display of multiple clusters of binding profiles. It uses standard Bioconductor classes for maximum compatibility with other software. Moreover, Starr automatically updates microarray probe annotation files by a highly efficient remapping of microarray probe sequences to an arbitrary genome. CONCLUSION: Starr is an R package that covers the complete ChIP-chip workflow from data processing to binding pattern detection. It focuses on the high-level data analysis, e.g., it provides methods for the integration and combined statistical analysis of binding profiles and complementary functional genomics data. Starr enables systematic assessment of binding behaviour for groups of genes that are alingned along arbitrary genomic features.