Since its selection as the method of the year in 2013, single-cell technologies have become mature enough to provide answers to complex research questions. With the growth of single-cell profiling technologies, there has also been a significant increase in data collected from single-cell profilings, resulting in computational challenges to process these massive and complicated datasets. To address these challenges, deep learning (DL) is positioned as a competitive alternative for single-cell analyses besides the traditional machine learning approaches. Here, we survey a total of 25 DL algorithms and their applicability for a specific step in the single cell RNA-seq processing pipeline. Specifically, we establish a unified mathematical representation of variational autoencoder, autoencoder, generative adversarial network and supervised DL models, compare the training strategies and loss functions for these models, and relate the loss functions of these models to specific objectives of the data processing step. Such a presentation will allow readers to choose suitable algorithms for their particular objective at each step in the pipeline. We envision that this survey will serve as an important information portal for learning the application of DL for scRNA-seq analysis and inspire innovative uses of DL to address a broader range of new challenges in emerging multi-omics and spatial single-cell sequencing.

BACKGROUND: The state-of-the-art deep learning based cancer type prediction can only predict cancer types whose samples are available during the training where the sample size is commonly large. In this paper, we consider how to utilize the existing training samples to predict cancer types unseen during the training. We hypothesize the existence of a set of type-agnostic expression representations that define the similarity/dissimilarity between samples of the same/different types and propose a novel one-shot learning model called CancerSiamese to learn this common representation. CancerSiamese accepts a pair of query and support samples (gene expression profiles) and learns the representation of similar or dissimilar cancer types through two parallel convolutional neural networks joined by a similarity function. RESULTS: We trained CancerSiamese for cancer type prediction for primary and metastatic tumors using samples from the Cancer Genome Atlas (TCGA) and MET500. Network transfer learning was utilized to facilitate the training of the CancerSiamese models. CancerSiamese was tested for different N-way predictions and yielded an average accuracy improvement of 8% and 4% over the benchmark 1-Nearest Neighbor (1-NN) classifier for primary and metastatic tumors, respectively. Moreover, we applied the guided gradient saliency map and feature selection to CancerSiamese to examine 100 and 200 top marker-gene candidates for the prediction of primary and metastatic cancers, respectively. Functional analysis of these marker genes revealed several cancer related functions between primary and metastatic tumors. CONCLUSION: This work demonstrated, for the first time, the feasibility of predicting unseen cancer types whose samples are limited. Thus, it could inspire new and ingenious applications of one-shot and few-shot learning solutions for improving cancer diagnosis, prognostic, and our understanding of cancer.

The recent accumulation of cancer genomic data provides an opportunity to understand how a tumor's genomic characteristics can affect its responses to drugs. This field, called pharmacogenomics, is a key area in the development of precision oncology. Deep learning (DL) methodology has emerged as a powerful technique to characterize and learn from rapidly accumulating pharmacogenomics data. We introduce the fundamentals and typical model architectures of DL. We review the use of DL in classification of cancers and cancer subtypes (diagnosis and treatment stratification of patients), prediction of drug response and drug synergy for individual tumors (treatment prioritization for a patient), drug repositioning and discovery and the study of mechanism/mode of action of treatments. For each topic, we summarize current genomics and pharmacogenomics data resources such as pan-cancer genomics data for cancer cell lines (CCLs) and tumors, and systematic pharmacologic screens of CCLs. By revisiting the published literature, including our in-house analyses, we demonstrate the unprecedented capability of DL enabled by rapid accumulation of data resources to decipher complex drug response patterns, thus potentially improving cancer medicine. Overall, this review provides an in-depth summary of state-of-the-art DL methods and up-to-date pharmacogenomics resources and future opportunities and challenges to realize the goal of precision oncology.

N6-methyladenosine (m6A) is the most abundant methylation, existing in >25% of human mRNAs. Exciting recent discoveries indicate the close involvement of m6A in regulating many different aspects of mRNA metabolism and diseases like cancer. However, our current knowledge about how m6A levels are controlled and whether and how regulation of m6A levels of a specific gene can play a role in cancer and other diseases is mostly elusive. We propose in this paper a computational scheme for predicting m6A-regulated genes and m6A-associated disease, which includes Deep-m6A, the first model for detecting condition-specific m6A sites from MeRIP-Seq data with a single base resolution using deep learning and Hot-m6A, a new network-based pipeline that prioritizes functional significant m6A genes and its associated diseases using the Protein-Protein Interaction (PPI) and gene-disease heterogeneous networks. We applied Deep-m6A and this pipeline to 75 MeRIP-seq human samples, which produced a compact set of 709 functionally significant m6A-regulated genes and nine functionally enriched subnetworks. The functional enrichment analysis of these genes and networks reveal that m6A targets key genes of many critical biological processes including transcription, cell organization and transport, and cell proliferation and cancer-related pathways such as Wnt pathway. The m6A-associated disease analysis prioritized five significantly associated diseases including leukemia and renal cell carcinoma. These results demonstrate the power of our proposed computational scheme and provide new leads for understanding m6A regulatory functions and its roles in diseases.

Gene regulatory networks are a global representation of complex interactions between molecules that dictate cellular behavior. Study of a regulatory network modulated by single or multiple modulators' expression levels, including microRNAs (miRNAs) and transcription factors (TFs), in different conditions can further reveal the modulators' roles in diseases such as cancers. Existing computational methods for identifying such modulated regulatory networks are typically carried out by comparing groups of samples dichotomized with respect to the modulator status, ignoring the fact that most biological features are intrinsically continuous variables. Here, we devised a sliding window-based regression scheme and proposed the Regression-based Inference of Modulation (RIM) algorithm to infer the dynamic gene regulation modulated by continuous-state modulators. We demonstrated the improvement in performance as well as computation efficiency achieved by RIM. Applying RIM to genome-wide expression profiles of 520 glioblastoma multiforme (GBM) tumors, we investigated miRNA- and TF-modulated gene regulatory networks and showed their association with dynamic cellular processes and brain-related functions in GBM. Overall, the proposed algorithm provides an efficient and robust scheme for comprehensively studying modulated gene regulatory networks.

N6-Methyladenosine (m6A) transcriptome methylation is an exciting new research area that just captures the attention of research community. We present in this paper, MeTDiff, a novel computational tool for predicting differential m6A methylation sites from Methylated RNA immunoprecipitation sequencing (MeRIP-Seq) data. Compared with the existing algorithm exomePeak, the advantages of MeTDiff are that it explicitly models the reads variation in data and also devices a more power likelihood ratio test for differential methylation site prediction. Comprehensive evaluation of MeTDiff's performance using both simulated and real datasets showed that MeTDiff is much more robust and achieved much higher sensitivity and specificity over exomePeak.

BACKGROUND: Recent studies illuminated a novel role of microRNA (miRNA) in the competing endogenous RNA (ceRNA) interaction: two genes (ceRNAs) can achieve coexpression by competing for a pool of common targeting miRNAs. Individual biological investigations implied ceRNA interaction performs crucial oncogenic/tumor suppressive functions in glioblastoma multiforme (GBM). Yet, a systematic analysis has not been conducted to explore the functional landscape and prognostic significance of ceRNA interaction. RESULTS: Incorporating the knowledge that ceRNA interaction is highly condition-specific and modulated by the expressional abundance of miRNAs, we devised a ceRNA inference by differential correlation analysis to identify the miRNA-modulated ceRNA pairs. Analyzing sample-paired miRNA and gene expression profiles of GBM, our data showed that this alternative layer of gene interaction is essential in global information flow. Functional annotation analysis revealed its involvement in activated processes in brain, such as synaptic transmission, as well as critical tumor-associated functions. Notably, a systematic survival analysis suggested the strength of ceRNA-ceRNA interactions, rather than expressional abundance of individual ceRNAs, among three immune response genes (CCL22, IL2RB, and IRF4) is predictive of patient survival. The prognostic value was validated in two independent cohorts. CONCLUSIONS: This work addresses the lack of a comprehensive exploration into the functional and prognostic relevance of ceRNA interaction in GBM. The proposed efficient and reliable method revealed its significance in GBM-related functions and prognosis. The highlighted roles of ceRNA interaction provide a basis for further biological and clinical investigations.

BACKGROUND: RNA sequencing (RNA-seq) is a high throughput technology that profiles gene expression in a genome-wide manner. RNA-seq has been mainly used for testing differential expression (DE) of transcripts between two conditions and has recently been used for testing differential alternative polyadenylation (APA). In the past, many algorithms have been developed for detecting differentially expressed genes (DEGs) from RNA-seq experiments, including the one we developed, XBSeq, which paid special attention to the context-specific background noise that is ignored in conventional gene expression quantification and DE analysis of RNA-seq data. RESULTS: We present several major updates in XBSeq2, including alternative statistical testing and parameter estimation method for detecting DEGs, capacity to directly process alignment files and methods for testing differential APA usage. We evaluated the performance of XBSeq2 against several other methods by using simulated datasets in terms of area under the receiver operating characteristic (ROC) curve (AUC), number of false discoveries and statistical power. We also benchmarked different methods concerning execution time and computational memory consumed. Finally, we demonstrated the functionality of XBSeq2 by using a set of in-house generated clear cell renal carcinoma (ccRCC) samples. CONCLUSIONS: We present several major updates to XBSeq. By using simulated datasets, we demonstrated that, overall, XBSeq2 performs equally well as XBSeq in terms of several statistical metrics and both perform better than DESeq2 and edgeR. In addition, XBSeq2 is faster in speed and consumes much less computational memory compared to XBSeq, allowing users to evaluate differential expression and APA events in parallel. XBSeq2 is available from Bioconductor: http://bioconductor.org/packages/XBSeq/.

Different types of genomic aberration may simultaneously contribute to tumorigenesis. To obtain a more accurate prognostic assessment to guide therapeutic regimen choice for cancer patients, the heterogeneous multi-omics data should be integrated harmoniously, which can often be difficult. For this purpose, we propose a Gene Interaction Regularized Elastic Net (GIREN) model that predicts clinical outcome by integrating multiple data types. GIREN conveniently embraces both gene measurements and gene-gene interaction information under an elastic net formulation, enforcing structure sparsity, and the "grouping effect" in solution to select the discriminate features with prognostic value. An iterative gradient descent algorithm is also developed to solve the model with regularized optimization. GIREN was applied to human ovarian cancer and breast cancer datasets obtained from The Cancer Genome Atlas, respectively. Result shows that, the proposed GIREN algorithm obtained more accurate and robust performance over competing algorithms (LASSO, Elastic Net, and Semi-supervised PCA, with or without average pathway expression features) in predicting cancer progression on both two datasets in terms of median area under curve (AUC) and interquartile range (IQR), suggesting a promising direction for more effective integration of gene measurement and gene interaction information.

MOTIVATION: N(6)-methyl-adenosine (m(6)A) is the most prevalent mRNA methylation but precise prediction of its mRNA location is important for understanding its function. A recent sequencing technology, known as Methylated RNA Immunoprecipitation Sequencing technology (MeRIP-seq), has been developed for transcriptome-wide profiling of m(6)A. We previously developed a peak calling algorithm called exomePeak. However, exomePeak over-simplifies data characteristics and ignores the reads' variances among replicates or reads dependency across a site region. To further improve the performance, new model is needed to address these important issues of MeRIP-seq data. RESULTS: We propose a novel, graphical model-based peak calling method, MeTPeak, for transcriptome-wide detection of m(6)A sites from MeRIP-seq data. MeTPeak explicitly models read count of an m(6)A site and introduces a hierarchical layer of Beta variables to capture the variances and a Hidden Markov model to characterize the reads dependency across a site. In addition, we developed a constrained Newton's method and designed a log-barrier function to compute analytically intractable, positively constrained Beta parameters. We applied our algorithm to simulated and real biological datasets and demonstrated significant improvement in detection performance and robustness over exomePeak. Prediction results on publicly available MeRIP-seq datasets are also validated and shown to be able to recapitulate the known patterns of m(6)A, further validating the improved performance of MeTPeak. AVAILABILITY AND IMPLEMENTATION: The package 'MeTPeak' is implemented in R and C ++, and additional details are available at https://github.com/compgenomics/MeTPeak CONTACT: [email protected] or [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Gene regulatory networks are a global representation of complex interactions between molecules that dictate cellular behavior. Study of a regulatory network modulated by single or multiple modulators' expression levels, including microRNAs (miRNAs) and transcription factors (TFs), in different conditions can further reveal the modulators' roles in diseases such as cancers. Existing computational methods for identifying such modulated regulatory networks are typically carried out by comparing groups of samples dichotomized with respect to the modulator status, ignoring the fact that most biological features are intrinsically continuous variables. Here we devised a sliding window-based regression scheme and proposed the Regression-based Inference of Modulation (RIM) algorithm to infer the dynamic gene regulation modulated by continuous-state modulators. We demonstrated the improvement in performance as well as computation efficiency achieved by RIM. Applying RIM to genome-wide expression profiles of 520 glioblastoma multiforme (GBM) tumors, we investigated miRNA- and TF-modulated gene regulatory networks and showed their association with dynamic cellular processes and brain-related functions in GBM. Overall, the proposed algorithm provides an efficient and robust scheme for comprehensively studying modulated gene regulatory networks.

DNA methylation in promoter regions has long been considered an essential mechanism of transcriptional regulation, and it has been demonstrated to be involved in cell development, tumor progression and aging. Methyl-CpG binding domain-based capture followed by high throughput sequencing (MBDCap-seq) is widely used to examine DNA methylation pattern genome-wide. Current MBDCap-seq data analysis approaches focus on measurement of methylated CpG sequence reads, without considering genomic characteristics and tissue-specific context and their impact to the amount of methylated DNA measurement (signal) and background fluctuation (noise). Therefore, specific software needs to be developed to process MBDCap-seq datasets. Here we presented a novel algorithm, termed MBDDiff, implemented as an R package that is designed specifically for processing MBDCap-seq datasets. MBDDiff contains three modules: quality assessment of datasets and quantification of DNA methylation; determination of differential methylation of promoter regions; and visualization functionalities. Simulation studies were carried out to demonstrate the accuracy of MBDDiff algorithm in detecting differential methylation in promoter regions. We also tested functionalities of MBDDiff to a set of in-house prostate cancer samples and a set of public-domain triple negative breast cancer samples profiled with MBDCap-seq protocol, and demonstrated the capability of identifying differential methylation of promoter regions of genes that might contribute to cancer development and progression.

Technology of next generation sequencing to detect pathogens of sample can impact human health by revealing pathogens which cause disease. Several workflow has developed in purposed to detect pathogens in next generation sequencing data. However, the requirement of computation power of these workflows limited the application. The time consuming problem make the workflow difficult to detect datasets with large sample size. Here we presented Patho-finder, a fast and accurate workflow designed for detecting pathogen in RNA sequencing data. We have evaluated performance of Patho-finder by three aspects. First, we evaluate performance by alter the data features, to see how Patho-finder work under different simulation conditions. Next, we compare the time consuming and accuracy between Patho-finder and existing workflow. At last, we used Patho-finder on the RNA-seq of cell lines with known virus-infected. The validation result demonstrated our approach could finish the task in real datasets.

MOTIVATION: When identifying differentially expressed (DE) genes from high-throughput gene expression measurements, we would like to take both statistical significance (such as P-value) and biological relevance (such as fold change) into consideration. In gene set enrichment analysis (GSEA), a score that can combine fold change and P-value together is needed for better gene ranking. RESULTS: We defined a gene significance score π-value by combining expression fold change and statistical significance (P-value), and explored its statistical properties. When compared to various existing methods, π-value based approach is more robust in selecting DE genes, with the largest area under curve in its receiver operating characteristic curve. We applied π-value to GSEA and found it comparable to P-value and t-statistic based methods, with added protection against false discovery in certain situations. Finally, in a gene functional study of breast cancer profiles, we showed that using π-value helps elucidating otherwise overlooked important biological functions. AVAILABILITY: http://gccri.uthscsa.edu/Pi_Value_Supplementary.asp CONTACT: [email protected], [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

DNA methylation is a common epigenetic marker that regulates gene expression. A robust and cost-effective way for measuring whole genome methylation is Methyl-CpG binding domain-based capture followed by sequencing (MBDCap-seq). In this study, we proposed BIMMER, a Hidden Markov Model (HMM) for differential Methylation Regions (DMRs) identification, where HMMs were proposed to model the methylation status in normal and cancer samples in the first layer and another HMM was introduced to model the relationship between differential methylation and methylation statuses in normal and cancer samples. To carry out the prediction for BIMMER, an Expectation-Maximization algorithm was derived. BIMMER was validated on the simulated data and applied to real MBDCap-seq data of normal and cancer samples. BIMMER revealed that 8.83% of the breast cancer genome are differentially methylated and the majority are hypo-methylated in breast cancer.

The articles in this special section were presented at the 2012 IEEE International Workshop on Genomic Signal Processing and Statistics (GENSIPS 2012) that was held in Washington DC from December 2nd to 4th.

Recent studies postulated that genes harboring identical microRNA (miRNA) binding sites can crosstalk by competing for a limited pool of the binding miRNAs (the miRNA program), named as the regulation of competing endogenous RNAs (ceRNAs). Incorporating recent biological evidence that ceRNA regulation depends on miRNA program expression levels, we developed, in the present study, a mathematical model for systematically inferring ceRNA regulation that is dependent on expression levels of the miRNA programs from sample-paired mRNA and miRNA expression datasets. Applying the method to analyze glioblastoma datasets, a compact ceRNA regulatory network was constructed. Our data further demonstrated that ceRNA regulation plays an essential role in transient cellular responses to dynamic inter-cellular signals. The findings illuminate mechanism of ceRNA regulation and further provide biological insights into the complex human interactome.

Methyl-CpG binding domain-based capture followed by sequencing (MBDCap-seq) is a cost-effective method for genome-wide methylation analyses especially in CpG-rich regions. In this study, we developed BIMMER, a BI-layer hidden Markov model for differential Methylation Regions (DMRs) identification BIMMER using MBDCap-seq samples derived from two different phenotypes. BIMMER models and generates a posterior probability for a 100bp bin to be a methylation site in either normal or disease samples by its first hidden layer, and then integrate these posterior probabilities in the second hidden layer to obtain the posterior probability of bin-specific differential methylation between the normal and disease samples. Based on these posterior probabilities, the decisions on the methylation and differential statuses for each bin can be calculated. Simulated results showed 94.3% area under precision-recall curve for BIMMER (BIMMER is programmed in Java and available by request).

Despite the prevalent studies of DNA/Chromatin related epigenetics, such as, histone modifications and DNA methylation, RNA epigenetics did not receive deserved attention due to the lack of high throughput approach for profiling epitranscriptome. Recently, a new affinity-based sequencing approach MeRIPseq was developed and applied to survey the global mRNA N6-methyladenosine (m6A) in mammalian cells. As a marriage of ChIPseq and RNAseq, MeRIPseq has the potential to study, for the first time, the transcriptome-wide distribution of different types of post-transcriptional RNA modifications. Yet, this technology introduced new computational challenges that have not been adequately addressed. We have previously developed a MATLAB-based package ‘exomePeak’ for detection of RNA methylation sites from MeRIPseq data. Here, we extend the features of exomePeak by including a novel computational framework that enables differential analysis to unveil the dynamics in RNA epigenetic regulations. The novel differential analysis monitors the percentage of modified RNA molecules among the total transcribed RNAs, which directly reflects the impact of RNA epigenetic regulations. In contrast, current available software packages developed for sequencing-based differential analysis such as DESeq or edgeR monitors the changes in the absolute amount of molecules, and, if applied to MeRIPseq data, might be dominated by transcriptional gene differential expression. The algorithm is implemented as an R-package ‘exomePeak’ and freely available. It takes directly the aligned BAM files as input, statistically supports biological replicates, corrects PCR artifacts, and outputs exome-based results in BED format, which is compatible with all major genome browsers for convenient visualization and manipulation. Examples are also provided to depict how exomePeak R-package is integrated with exiting tools for MeRIPseq based peak calling and differential analysis. Particularly, the rationales behind each processing step as well as the specific method used, the best practice, and possible alternative strategies are briefly discussed. The algorithm was applied to the human HepG2 cell MeRIPseq data sets and detects more than 16000 RNA m6A sites, many of which are differentially methylated under ultraviolet radiation. The challenges and potentials of MeRIPseq in epitranscriptome studies are discussed in the end.

Integrative clinical outcome prediction model called gene interaction regularized elastic net (GIREN) method is proposed in this paper. GIREN combines gene expression, methylation profiles, and gene interaction networks in order to reveal genomic and epigenomic features that bear important prognostic value. With GIREN, gene expression and DNA methylation profiles are first jointly analyzed in a linear regression model, and additional gene interaction network is simultaneously integrated as a regularizing penalty that follow an elastic net formulation. Such regularization also enforce sparsity in the solution so that features with prognostic values are automatically selected. To solve the regularized optimization, an iterative gradient descent algorithm is also developed. We applied GIREN to a set of 87 human ovarian cancer samples, which underwent a rigorous sample selection. The predicted outcome was used to group patients into high-risk vs. low-risk. Validation showed that GIREN outperformed other competing algorithms including SuperPCA.

MOTIVATION: Fragmented RNA immunoprecipitation combined with RNA sequencing enabled the unbiased study of RNA epigenome at a near single-base resolution; however, unique features of this new type of data call for novel computational techniques. RESULT: Through examining the connections of RNA epigenome sequencing data with two well-studied data types, ChIP-Seq and RNA-Seq, we unveiled the salient characteristics of this new data type. The computational strategies were discussed accordingly, and a novel data processing pipeline was proposed that combines several existing tools with a newly developed exome-based approach 'exomePeak' for detecting, representing and visualizing the post-transcriptional RNA modification sites on the transcriptome. AVAILABILITY: The MATLAB package 'exomePeak' and additional details are available at http://compgenomics.utsa.edu/exomePeak/.

The number of current anti-cancer drugs was limited and the response rates were also not high. To "reposition" known drugs as anti-cancer drugs to increase the therapeutic efficiency, we presented a novel analysis framework to identify putative drugs for cancer. Using breast cancer as example, a "cancer - gene sets - drugs" network was constructed through two procedures. First, the "gene sets - drugs" network was built by applying the expression pattern of drugs for gene set enrichment analysis. Secondly, the breast cancer progression associated gene sets were identified by survival analysis of patient cohorts. By integrating the two results, 25 tumor progression associated gene sets and 360 putative anti-cancer drugs were identified. Our method has the ability to identify the "reposition" drugs and the potential affected mechanisms of tumor progression concurrently. It will be useful to speed up the development of anti-cancer drugs from bench to clinical application.

Transcriptional regulation by transcription factors (TFs) and microRNAs controls when and how much RNA is created. Due to technical limitations, the protein level expressions of TFs are usually unknown, making computational reconstruction of transcriptional network a difficult task. We proposed here a novel Bayesian non negative hybrid factor model for transcriptional network modeling, which is capable to estimate both the non-negative abundances of the transcription factors, the regulatory effects of TFs and microRNAs, and the sample clustering information by integrating microarray data and existing knowledge regarding TFs and microRNAs regulated target genes. The results demonstrated its validity and effectiveness to reconstructing transcriptional networks through simulated systems and real data.

The problem of uncovering transcriptional regulation by transcription factors (TFs) based on microarray data is considered. A novel Bayesian sparse correlated rectified factor model (BSCRFM) coupled with its ICM solution is proposed. BSCRFM models the unknown TF protein level activity, the correlated regulations between TFs, and the sparse nature of TF regulated genes and it admits prior knowledge from existing database regarding TF regulated target genes. An efficient Iterated Conditional Modes (ICM) algorithm is developed, and a maximum a posterior (MAP) solution is calculated from multiple ICM results to avoid the local maximum problem, a context-specific transcriptional regulatory network specific to the experimental condition of the microarray data can then be obtained. The proposed model's ICM algorithm and MAP solution are evaluated on the simulated systems and results demonstrated the validity and effectiveness of the proposed approach. The proposed model is also applied to the breast cancer microarray data and a TF regulated network is obtained.

BACKGROUND: MicroRNAs (miRNAs) are single-stranded non-coding RNAs known to regulate a wide range of cellular processes by silencing the gene expression at the protein and/or mRNA levels. Computational prediction of miRNA targets is essential for elucidating the detailed functions of miRNA. However, the prediction specificity and sensitivity of the existing algorithms are still poor to generate meaningful, workable hypotheses for subsequent experimental testing. Constructing a richer and more reliable training data set and developing an algorithm that properly exploits this data set would be the key to improve the performance current prediction algorithms. RESULTS: A comprehensive training data set is constructed for mammalian miRNAs with its positive targets obtained from the most up-to-date miRNA target depository called miRecords and its negative targets derived from 20 microarray data. A new algorithm SVMicrO is developed, which assumes a 2-stage structure including a site support vector machine (SVM) followed by a UTR-SVM. SVMicrO makes prediction based on 21 optimal site features and 18 optimal UTR features, selected by training from a comprehensive collection of 113 site and 30 UTR features. Comprehensive evaluation of SVMicrO performance has been carried out on the training data, proteomics data, and immunoprecipitation (IP) pull-down data. Comparisons with some popular algorithms demonstrate consistent improvements in prediction specificity, sensitivity and precision in all tested cases. All the related materials including source code and genome-wide prediction of human targets are available at http://compgenomics.utsa.edu/svmicro.html. CONCLUSIONS: A 2-stage SVM based new miRNA target prediction algorithm called SVMicrO is developed. SVMicrO is shown to be able to achieve robust performance. It holds the promise to achieve continuing improvement whenever better training data that contain additional verified or high confidence positive targets and properly selected negative targets are available.

MOTIVATION: Genomic instability is one of the fundamental factors in tumorigenesis and tumor progression. Many studies have shown that copy-number abnormalities at the DNA level are important in the pathogenesis of cancer. Array comparative genomic hybridization (aCGH), developed based on expression microarray technology, can reveal the chromosomal aberrations in segmental copies at a high resolution. However, due to the nature of aCGH, many standard expression data processing tools, such as data normalization, often fail to yield satisfactory results. RESULTS: We demonstrated a novel aCGH normalization algorithm, which provides an accurate aCGH data normalization by utilizing the dependency of neighboring probe measurements in aCGH experiments. To facilitate the study, we have developed a hidden Markov model (HMM) to simulate a series of aCGH experiments with random DNA copy number alterations that are used to validate the performance of our normalization. In addition, we applied the proposed normalization algorithm to an aCGH study of lung cancer cell lines. By using the proposed algorithm, data quality and the reliability of experimental results are significantly improved, and the distinct patterns of DNA copy number alternations are observed among those lung cancer cell lines. SUPPLEMENTARY INFORMATION: Source codes and.gures may be found at http://ntumaps.cgm.ntu.edu.tw/aCGH_supplementary.

Bioinformatics 2008; Vol. 24 no. 16: 1749–1756. The publishers regret that there was an error in the copyright line which should have been Open Access. This article is now freely available online. The copyright line of the article should have read: © 2008 The Author(s) This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/2.0/uk/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

UNLABELLED: The NCBI Gene Expression Omnibus (GEO) represents the largest public repository of microarray data. However, finding data in GEO can be challenging. We have developed GEOmetadb in an attempt to make querying the GEO metadata both easier and more powerful. All GEO metadata records as well as the relationships between them are parsed and stored in a local MySQL database. A powerful, flexible web search interface with several convenient utilities provides query capabilities not available via NCBI tools. In addition, a Bioconductor package, GEOmetadb that utilizes a SQLite export of the entire GEOmetadb database is also available, rendering the entire GEO database accessible with full power of SQL-based queries from within R. AVAILABILITY: The web interface and SQLite databases available at http://gbnci.abcc.ncifcrf.gov/geo/. The Bioconductor package is available via the Bioconductor project. The corresponding MATLAB implementation is also available at the same website.

MOTIVATION: There are two general methods for making gene-expression microarrays: one is to hybridize a single test set of labeled targets to the probe, and measure the background-subtracted intensity at each probe site; the other is to hybridize both a test and a reference set of differentially labeled targets to a single detector array, and measure the ratio of the background-subtracted intensities at each probe site. Which method is better depends on the variability in the cell system and the random factors resulting from the microarray technology. It also depends on the purpose for which the microarray is being used. Classification is a fundamental application and it is the one considered here. RESULTS: This paper describes a model-based simulation paradigm that compares the classification accuracy provided by these methods over a variety of noise types and presents the results of a study modeled on noise typical of cDNA microarray data. The model consists of four parts: (1) the measurement equation for genes in the reference state; (2) the measurement equation for genes in the test state; (3) the ratio and normalization procedure for a dual-channel system; and (4) the intensity and normalization procedure for a single-channel system. In the reference state, the mean intensities are modeled as a shifted exponential distribution, and the intensity for a particular gene is modeled via a normal distribution, Normal(I, alphaI), about its mean intensity I, with alpha being the coefficient of variation of the cell system. In the test state, some genes have their intensities up-regulated by a random factor. The model includes a number of random factors affecting intensity measurement: deposition gain d, labeling gain, and post-image-processing residual noise. The key conclusion resulting from the study is that the coefficient of variation governing the randomness of the intensities and the deposition gain are the most important factors for determining whether a single-channel or dual-channel system provides superior classification, and the decision region in the alpha-d plane is approximately linear.

MOTIVATION: Expression-based analysis for large families of genes has recently become possible owing to the development of cDNA microarrays, which allow simultaneous measurement of transcript levels for thousands of genes. For each spot on a microarray, signals in two channels must be extracted from their backgrounds. This requires algorithms to extract signals arising from tagged mRNA hybridized to arrayed cDNA locations and algorithms to determine the significance of signal ratios. RESULTS: This paper focuses on estimation of signal ratios from the two channels, and the significance of those ratios. The key issue is the determination of whether a ratio is significantly high or low in order to conclude whether the gene is upregulated or downregulated. The paper builds on an earlier study that involved a hypothesis test based on a ratio statistic under the supposition that the measured fluorescent intensities subsequent to image processing can be assumed to reflect the signal intensities. Here, a refined hypothesis test is considered in which the measured intensities forming the ratio are assumed to be combinations of signal and background. The new method involves a signal-to-noise ratio, and for a high signal-to-noise ratio the new test reduces (with close approximation) to the original test. The effect of low signal-to-noise ratio on the ratio statistics constitutes the main theme of the paper. Finally, and in this vein, a quality metric is formulated for spots. This measure can be used to decide whether or not a spot ratio should be deleted, or to adjust various measurements to reflect confidence in the quality of the measurement. CONTACT: [email protected]

MOTIVATION: The simplest level of statistical analysis of cancer associated gene expression matrices is aimed at finding consistently up- or down-regulated genes within a given set of tumor samples. Considering the high level of gene expression diversity detected in cancer, one needs to assess the probability that the consistent mis-regulation of a given gene is due to chance. Furthermore, it is important to determine the required sample number that will ensure the meaningful statistical analysis of massively parallel gene expression measurements. RESULTS: The probability of consistent mis-regulation is calculated in this paper for binarized gene expression data, using combinatorial considerations. For practical purposes, we also provide a set of accurate approximate formulas for determining the same probability in a computationally less intensive way. When the pool of mis-regulatable genes is restricted, the probability of consistent mis-regulation can be overestimated. We show, however, that this effect has little practical consequences for cancer associated gene expression measurements published in the literature. Finally, in order to aid experimental design, we have provided estimates on the required sample number that will ensure that the detected consistent mis-regulation is not due to chance. Our results suggest that less than 20 sufficiently diverse tumor samples may be enough to identify consistently mis-regulated genes in a statistically significant manner. AVAILABILITY: An implementation using Mathematica (tm) of the main equation of the paper, (4), is available at www.me.chalmers.se/~mwahde/bioinfo.html.