Search

Background Piwi-interacting RNAs (piRNAs) are the small non-coding RNAs (ncRNAs) that silence genomic transposable elements. And researchers found out that piRNA also regulates various endogenous transcripts. However, there is no systematic understanding of the piRNA binding patterns and how piRNA targets genes. While various prediction methods have been developed for other similar ncRNAs (e.g., miRNAs), piRNA holds distinctive characteristics and requires its own computational model for binding target prediction. Results Recently, transcriptome-wide piRNA binding events in C. elegans were probed by PRG-1 CLASH experiments. Based on the probed piRNA-messenger RNAs (mRNAs) binding pairs, in this research, we devised the first deep learning architecture based on multi-head attention to computationally identify piRNA targeting mRNA sites. In the devised deep network, the given piRNA and mRNA segment sequences are first one-hot encoded and undergo a combined operation of convolution and squeezing-extraction to unravel motif patterns. And we incorporate a novel multi-head attention sub-network to extract the hidden piRNA binding rules that can simulate the biological piRNA target recognition process. Finally, the true piRNA–mRNA binding pairs are identified by a deep fully connected sub-network. Our model obtains a supreme discriminatory power of AUC $$=$$ = 93.3% on an independent test set and successfully extracts the verified binding pattern of a synthetic piRNA. These results demonstrated that the devised model achieves high prediction performance and suggests testable potential biological piRNA binding rules. Conclusions In this research, we developed the first deep learning method to identify piRNA targeting sites on C. elegans mRNAs. And the developed deep learning method is demonstrated to be of high accuracy and can provide biological insights into piRNA–mRNA binding patterns. The piRNA binding target identification network can be downloaded from http://cosbi2.ee.ncku.edu.tw/data_download/piRNA_mRNA_binding.

Transcript isoforms regulated by alternative splicing can substantially impact carcinogenesis, leading to a need to obtain clues for both gene differential expression and malfunctions of isoform distributions in cancer studies. The Cancer Genome Atlas (TCGA) project was launched in 2008 to collect cancer-related genome mutation raw data from the population. While many repositories tried to add insights into the raw data in TCGA, no existing database provides both comprehensive gene-level and isoform-level cancer stage marker investigation and survival analysis. We constructed Cancer DEIso to facilitate in-depth analyses for both gene-level and isoform-level human cancer studies. Patient RNA-seq data, sample sheets, patient clinical data, and human genome datasets were collected and processed in Cancer DEIso. And four functions to search differentially expressed genes/isoforms between cancer stages were implemented: (i) Search potential gene/isoform markers for a specified cancer type and its two stages; (ii) Search potentially induced cancer types and stages for a gene/isoform; (iii) Expression survival analysis on a given gene/isoform for some cancer; (iv) Gene/isoform stage expression comparison visualization. As an example, we demonstrate that Cancer DEIso can indicate potential colorectal cancer isoform diagnostic markers that are not easily detected when only gene-level expressions are considered. Cancer DEIso is available at http://cosbi4.ee.ncku.edu.tw/DEIso/.

Background Translational regulation is one important aspect of gene expression regulation. Dysregulation of translation results in abnormal cell physiology and leads to diseases. Ribosome profiling (RP), also called ribo-seq, is a powerful experimental technique to study translational regulation. It can capture a snapshot of translation by deep sequencing of ribosome-protected mRNA fragments. Many ribosome profiling data processing tools have been developed. However, almost all tools analyze ribosome profiling data at the gene level. Since different isoforms of a gene may produce different proteins with distinct biological functions, it is advantageous to analyze ribosome profiling data at the isoform level. To meet this need, previously we developed a pipeline to analyze 610 public human ribosome profiling data at the isoform level and constructed HRPDviewer database. Results To allow other researchers to use our pipeline as well, here we implement our pipeline as an easy-to-use software tool called RPiso. Compared to Ribomap (a widely used tool which provides isoform-level ribosome profiling analyses), our RPiso (1) estimates isoform abundance more accurately, (2) supports analyses on more species, and (3) provides a web-based viewer for interactively visualizing ribosome profiling data on the selected mRNA isoforms. Conclusions In this study, we developed RPiso software tool (http://cosbi7.ee.ncku.edu.tw/RPiso/) to provide isoform-level ribosome profiling analyses. RPiso is very easy to install and execute. RPiso also provides a web-based viewer for interactively visualizing ribosome profiling data on the selected mRNA isoforms. We believe that RPiso is a useful tool for researchers to analyze and visualize their own ribosome profiling data at the isoform level.

Graphical abstract, Highlights • YPIBP contains 679 PI-binding yeast proteins and their PI-binding evidence. • YPIBP contains 181 lipid-binding domains and 103 lipid-related domains. • YPIBP provides PI-binding enrichment analysis and domain enrichment analysis. • YPIBP provides visualization of PI-domain-protein interactome., Phosphoinositides (PIs) are a family of eight lipids consisting of phosphatidylinositol (PtdIns) and its seven phosphorylated forms. PIs have important regulatory functions in the cell including lipid signaling, protein transport, and membrane trafficking. Yeast has been recognized as a eukaryotic model system to study lipid-protein interactions. Hundreds of yeast PI-binding proteins have been identified, but this research knowledge remains scattered. Besides, the complete PI-binding spectrum and potential PI-binding domains have not been interlinked. No comprehensive databases are available to support the lipid-protein interaction research on phosphoinositides. Here we constructed the first knowledgebase of Yeast Phosphoinositide-Binding Proteins (YPIBP), a repository consisting of 679 PI-binding proteins collected from high-throughput proteome-array and lipid-array studies, QuickGO, and a rigorous literature mining. The YPIBP also contains protein domain information in categories of lipid-binding domains, lipid-related domains and other domains. The YPIBP provides search and browse modes along with two enrichment analyses (PI-binding enrichment analysis and domain enrichment analysis). An interactive visualization is given to summarize the PI-domain-protein interactome. Finally, three case studies were given to demonstrate the utility of YPIBP. The YPIBP knowledgebase consolidates the present knowledge and provides new insights of the PI-binding proteins by bringing comprehensive and in-depth interaction network of the PI-binding proteins. YPIBP is available at http://cosbi7.ee.ncku.edu.tw/YPIBP/.

Nowadays high-throughput omics technologies are routinely used in biological research. From the omics data, researchers can easily get two gene lists (e.g. stress-induced genes vs. stress-repressed genes) related to their biological question. The next step would be to apply enrichment analysis tools to identify distinct functional/regulatory features between these two gene lists for further investigation. Although various enrichment analysis tools are already available, two challenges remain to be addressed. First, most existing tools are designed to analyze only one gene list, so they cannot directly compare two gene lists. Second, almost all existing tools focus on identifying the enriched qualitative features (e.g. gene ontology [GO] terms, pathways, domains, etc.). Many quantitative features (e.g. number of mRNA isoforms of a gene, mRNA half-life, protein half-life, transcriptional plasticity, translational efficiency, etc.) are available in the yeast, but no existing tools provide analyses on these quantitative features. To address these two challenges, here we present Yeast Quantitative Features Comparator (YQFC) that can directly compare various quantitative features between two yeast gene lists. In YQFC, we comprehensively collected and processed 85 quantitative features from the yeast literature and yeast databases. For each quantitative feature, YQFC provides three statistical tests (t-test, U test and KS test) to test whether this quantitative feature is statistically different between the two input yeast gene lists. The distinct quantitative features identified by YQFC may help researchers to study the underlying molecular mechanisms that differentiate the two input yeast gene lists. We believe that YQFC is a useful tool to expedite the biological research that uses high-throughput omics technologies.Database URLhttp://cosbi2.ee.ncku.edu.tw/YQFC/

Self-defense by avoiding self-targeting By silencing transposons, Piwi-interacting RNAs (piRNAs) protect the stability of animal genomes in germ lines. However, many piRNAs do not map to transposons, and their functions have remained undefined. Zhang et al. described the piRNA targeting logic in Caenorhabditis elegans and identified an intrinsic sequence signal in endogenous germline genes that confer resistance to piRNA silencing. Thus, diverse piRNAs silence foreign nucleic acids but spare self genes to defend the C. elegans genome. In addition, multiple foreign transgenes can be engineered to escape piRNA targeting, allowing successful expression in the germline. Science , this issue p. 587

Post-translational modifications of histones (e.g. acetylation, methylation, phosphorylation and ubiquitination) play crucial roles in regulating gene expression by altering chromatin structures and creating docking sites for histone/chromatin regulators. However, the combination patterns of histone modifications, regulatory proteins and their corresponding target genes remain incompletely understood. Therefore, it is advantageous to have a tool for the enrichment/depletion analysis of histone modifications and histone/chromatin regulators from a gene list. Many ChIP-chip/ChIP-seq datasets of histone modifications and histone/chromatin regulators in yeast can be found in the literature. Knowing the needs and having the data motivate us to develop a web tool, called Yeast Histone Modifications Identifier (YHMI), which can identify the enriched/depleted histone modifications and the enriched histone/chromatin regulators from a list of yeast genes. Both tables and figures are provided to visualize the identification results. Finally, the high-quality and biological insight of the identification results are demonstrated by two case studies. We believe that YHMI is a valuable tool for yeast biologists to do epigenetics research.

Arsenic is a toxic metalloid. Moderate levels of arsenic exposure from drinking water can cause various human health problems such as skin lesions, circulatory disorders and cancers. Thus, arsenic toxicity is a key focus area for environmental and toxicological investigations. Many arsenic-related genes in yeast have been identified by experimental strategies such as phenotypic screening and transcriptional profiling. These identified arsenic-related genes are valuable information for studying arsenic toxicity. However, the literature about these identified arsenic-related genes is widely dispersed and cannot be easily acquired by researchers. This prompts us to develop YARG (Yeast Arsenic-Related Genes) database, which comprehensively collects 3396 arsenic-related genes in the literature. For each arsenic-related gene, the number and types of experimental evidence (phenotypic screening and/or transcriptional profiling) are provided. Users can use both search and browse modes to query arsenic-related genes in YARG. We used two case studies to show that YARG can return biologically meaningful arsenic-related information for the query gene(s). We believe that YARG is a useful resource for arsenic toxicity research. YARG is available at http://cosbi4.ee.ncku.edu.tw/YARG/.

Translational regulation plays an important role in protein synthesis. Dysregulation of translation causes abnormal cell physiology and leads to diseases such as inflammatory disorders and cancers. An emerging technique, called ribosome profiling (ribo-seq), was developed to capture a snapshot of translation. It is based on deep sequencing of ribosome-protected mRNA fragments. A lot of ribo-seq data have been generated in various studies, so databases are needed for depositing and visualizing the published ribo-seq data. Nowadays, GWIPS-viz, RPFdb and TranslatomeDB are the three largest databases developed for this purpose. However, two challenges remain to be addressed. First, GWIPS-viz and RPFdb databases align the published ribo-seq data to the genome. Since ribo-seq data aim to reveal the actively translated mRNA transcripts, there are advantages of aligning ribo-req data to the transcriptome over the genome. Second, TranslatomeDB does not provide any visualization and the other two databases only provide visualization of the ribo-seq data around a specific genomic location, while simultaneous visualization of the ribo-seq data on multiple mRNA transcripts produced from the same gene or different genes is desired. To address these two challenges, we developed the Human Ribosome Profiling Data viewer (HRPDviewer). HRPDviewer (i) contains 610 published human ribo-seq datasets from Gene Expression Omnibus, (ii) aligns the ribo-seq data to the transcriptome and (iii) provides visualization of the ribo-seq data on the selected mRNA transcripts. Using HRPDviewer, researchers can compare the ribosome binding patterns of multiple mRNA transcripts from the same gene or different genes to gain an accurate understanding of protein synthesis in human cells. We believe that HRPDviewer is a useful resource for researchers to study translational regulation in human. Database URL: http://cosbi4.ee.ncku.edu.tw/HRPDviewer/ or http://cosbi5.ee.ncku.edu.tw/HRPDviewer/

pirScan is a web-based tool for identifying C. elegans piRNA-targeting sites within a given mRNA or spliced DNA sequence. The purpose of our tool is to allow C. elegans researchers to predict piRNA targeting sites and to avoid the persistent germline silencing of transgenes that has rendered many constructs unusable. pirScan fulfills this purpose by first enumerating the predicted piRNA-targeting sites present in an input sequence. This prediction can be exported in a tabular or graphical format. Subsequently, pirScan suggests silent mutations that can be introduced to the input sequence that would allow the modified transgene to avoid piRNA targeting. The user can customize the piRNA targeting stringency and the silent mutations that he/she wants to introduce into the sequence. The modified sequences can be re-submitted to be certain that any previously present piRNA-targeting sites are now absent and no new piRNA-targeting sites are accidentally generated. This revised sequence can finally be downloaded as a text file and/or visualized in a graphical format. pirScan is freely available for academic use at http://cosbi4.ee.ncku.edu.tw/pirScan/.

By organizing the genome into gene modules (GMs), a living cell coordinates the activities of a set of genes to properly respond to environmental changes. The transcriptional regulation of the expression of a GM is usually carried out by a cooperative transcription factor set (CoopTFS) consisting of several cooperative transcription factors (TFs). Therefore, a database which provides CoopTFSs and their target GMs is useful for studying the cellular responses to internal or external stimuli. To address this need, here we constructed YGMD (Yeast Gene Module Database) to provide 34120 CoopTFSs, each of which consists of two to five cooperative TFs, and their target GMs. The cooperativity between TFs in a CoopTFS is suggested by physical/genetic interaction evidence or/and predicted by existing algorithms. The target GM regulated by a CoopTFS is defined as the common target genes of all the TFs in that CoopTFS. The regulatory association between any TF in a CoopTFS and any gene in the target GM is supported by experimental evidence in the literature. In YGMD, users can (i) search the GM regulated by a specific CoopTFS of interest or (ii) search all possible CoopTFSs whose target GMs contain a specific gene of interest. The biological relevance of YGMD is shown by a case study which demonstrates that YGMD can provide a GM enriched with genes known to be regulated by the query CoopTFS (Cbf1-Met4-Met32). We believe that YGMD provides a valuable resource for yeast biologists to study the transcriptional regulation of GMs. Database URL: http://cosbi4.ee.ncku.edu.tw/YGMD/, http://cosbi5.ee.ncku.edu.tw/YGMD/ or http://cosbi.ee.ncku.edu.tw/YGMD/

Background Missing value imputation is important for microarray data analyses because microarray data with missing values would significantly degrade the performance of the downstream analyses. Although many microarray missing value imputation algorithms have been developed, an objective and comprehensive performance comparison framework is still lacking. To solve this problem, we previously proposed a framework which can perform a comprehensive performance comparison of different existing algorithms. Also the performance of a new algorithm can be evaluated by our performance comparison framework. However, constructing our framework is not an easy task for the interested researchers. To save researchers’ time and efforts, here we present an easy-to-use web tool named MVIAeval (Missing Value Imputation Algorithm evaluator) which implements our performance comparison framework. Results MVIAeval provides a user-friendly interface allowing users to upload the R code of their new algorithm and select (i) the test datasets among 20 benchmark microarray (time series and non-time series) datasets, (ii) the compared algorithms among 12 existing algorithms, (iii) the performance indices from three existing ones, (iv) the comprehensive performance scores from two possible choices, and (v) the number of simulation runs. The comprehensive performance comparison results are then generated and shown as both figures and tables. Conclusions MVIAeval is a useful tool for researchers to easily conduct a comprehensive and objective performance evaluation of their newly developed missing value imputation algorithm for microarray data or any data which can be represented as a matrix form (e.g. NGS data or proteomics data). Thus, MVIAeval will greatly expedite the progress in the research of missing value imputation algorithms.

Identifying cell cycle transcription factors (TFs) is important for understanding the transcriptional regulation of the cell cycle process which controls the growth and development of all organisms. Existing computational approaches for identifying cell cycle TFs are mainly based on methods with a fixed selection criterion. That is, the same criterion was applied to each TF to determine whether it is a cell cycle TF or not. Since the characteristic of each TF may be quite different, it is not suitable to use a fixed selection criterion in identifying cell cycle TFs. Instead of using a fixed selection criterion, we propose a method with variable selection criteria to identify cell cycle TFs in yeast by integrating the ChIP-chip and cell cycle gene expression data. Our method is shown to outperform five existing methods which used the same ChIP-chip dataset as we did. Fifteen cell cycle TFs were identified by our approach, 12 of which are known cell cycle TFs, while the remaining three (Hap4, Reb1 and Tye7) are novel cell cycle TFs. The biological significance of our predictions is shown by four lines of indirect evidence derived from the protein-protein interaction data, TF mutant data, ChIP-chip data and the results of the previous computational studies.

Transcription factors control gene expression by binding to short specific DNA sequences, called transcription factor binding sites (TFBSs), in the promoter of a gene. Thus, studying the spatial distribution of TFBSs in the promoters may provide insights into the molecular mechanisms of gene regulation. I developed a method to construct the spatial distribution of TFBSs for any set of genes of interest. I found that different functional gene clusters have different spatial distributions of TFBSs, indicating that gene regulation mechanisms may be very different among different functional gene clusters. I also found that the binding sites for different transcription factors (TFs) may have different spatial distributions: a sharp peak, a plateau or no dominant single peak. The spatial distributions of binding sites for many TFs derived from my analyses are valuable prior information for TFBS prediction algorithm because different regions of a promoter can assign different possibilities for TFBS occurrence.

ChIP-Chip data, which shows binding of transcription factors (TFs) to promoter regions in vivo, are widely used by biologists to identify the regulatory targets of TFs. However, the binding of a TF to a gene does not necessarily imply regulation. Thus, it is important to develop computational methods which can extract a TF's regulatory targets from its binding targets. We developed a method, called REgulatory Targets Extraction Algorithm (RETEA), which uses partial correlation analysis on gene expression data to extract a TF's regulatory targets from its binding targets inferred from ChIP-chip data. We applied RETEA to yeast cell cycle microarray data and identified the plausible regulatory targets of eleven known cell cycle TFs. We validated our predictions by checking the enrichments for cell cycle-regulated genes, common cellular processes and common molecular functions. Finally, we showed that RETEA performs better than three published methods (MA-Network, TRIA and Garten et al's method).

Unicellular organisms such as yeasts have evolved mechanisms to respond to environmental stresses by rapidly reorganizing the gene expression program. Although many stress-response genes in yeast have been discovered by DNA microarrays, the stress-response transcription factors (TFs) that regulate these stress-response genes remain to be investigated. In this study, we use a dynamic system model of gene regulation to describe the mechanism of how TFs may control a gene's expression. Then, based on the dynamic system model, we develop the Stress Regulator Identification Algorithm (SRIA) to identify stress-response TFs for six kinds of stresses. We identified some general stress-response TFs that respond to various stresses and some specific stress-response TFs that respond to one specific stress. The biological significance of our findings is validated by the literature. We found that a small number of TFs is probably sufficient to control a wide variety of expression patterns in yeast under different stresses. Two implications can be inferred from this observation. First, the response mechanisms to different stresses may have a bow-tie structure. Second, there may be regulatory cross-talks among different stress responses. In conclusion, this study proposes a network of stress-response regulators and the details of their actions.

The eukaryotic cell cycle is a complex process and is precisely regulated at many levels. Many genes specific to the cell cycle are regulated transcriptionally and are expressed just before they are needed. To understand the cell cycle process, it is important to identify the cell cycle transcription factors (TFs) that regulate the expression of cell cycle-regulated genes. Here, we describe a computational method to identify cell cycle TFs in yeast by integrating current ChIP-chip, mutant, transcription factor-binding site (TFBS), and cell cycle gene expression data. For each identified cell cycle TF, our method also assigned specific cell cycle phases in which the TF functions and identified the time lag for the TF to exert regulatory effects on its target genes. Moreover, our method can identify novel cell cycle-regulated genes as a by-product.

Background Eukaryotic transcriptional regulation is known to be highly connected through the networks of cooperative transcription factors (TFs). Measuring the cooperativity of TFs is helpful for understanding the biological relevance of these TFs in regulating genes. The recent advances in computational techniques led to various predictions of cooperative TF pairs in yeast. As each algorithm integrated different data resources and was developed based on different rationales, it possessed its own merit and claimed outperforming others. However, the claim was prone to subjectivity because each algorithm compared with only a few other algorithms and only used a small set of performance indices for comparison. This motivated us to propose a series of indices to objectively evaluate the prediction performance of existing algorithms. And based on the proposed performance indices, we conducted a comprehensive performance evaluation. Results We collected 14 sets of predicted cooperative TF pairs (PCTFPs) in yeast from 14 existing algorithms in the literature. Using the eight performance indices we adopted/proposed, the cooperativity of each PCTFP was measured and a ranking score according to the mean cooperativity of the set was given to each set of PCTFPs under evaluation for each performance index. It was seen that the ranking scores of a set of PCTFPs vary with different performance indices, implying that an algorithm used in predicting cooperative TF pairs is of strength somewhere but may be of weakness elsewhere. We finally made a comprehensive ranking for these 14 sets. The results showed that Wang J's study obtained the best performance evaluation on the prediction of cooperative TF pairs in yeast. Conclusions In this study, we adopted/proposed eight performance indices to make a comprehensive performance evaluation on the prediction results of 14 existing cooperative TFs identification algorithms. Most importantly, these proposed indices can be easily applied to measure the performance of new algorithms developed in the future, thus expedite progress in this research field.

Background Defining a measure for regulatory similarity (RS) of two genes is an important step toward identifying co-regulated genes. To date, transcription factor binding sites (TFBSs) have been widely used to measure the RS of two genes because transcription factors (TFs) binding to TFBSs in promoters is the most crucial and well understood step in gene regulation. However, existing TFBS-based RS measures consider the relation of a TFBS to a gene as a Boolean (either 'presence' or 'absence') without utilizing the information of TFBS locations in promoters. Results Functional TFBSs of many TFs in yeast are known to have a strong positional preference to occur in a small region in the promoters. This biological knowledge prompts us to develop a novel RS measure that exploits the TFBS location information. The performances of different RS measures are evaluated by the fraction of gene pairs that are co-regulated (validated by literature evidence) by at least one common TF under different RS scores. The experimental results show that the proposed RS measure is the best co-regulation indicator among the six compared RS measures. In addition, the co-regulated genes identified by the proposed RS measure are also shown to be able to benefit three co-regulation-based applications: detecting gene co-function, gene co-expression and protein-protein interactions. Conclusions The proposed RS measure provides a good indicator for gene co-regulation. Besides, its good performance reveals the importance of the location information in TFBS-based RS measures.

Background Cis-regulatory modules (CRMs), or the DNA sequences required for regulating gene expression, play the central role in biological researches on transcriptional regulation in metazoan species. Nowadays, the systematic understanding of CRMs still mainly resorts to computational methods due to the time-consuming and small-scale nature of experimental methods. But the accuracy and reliability of different CRM prediction tools are still unclear. Without comparative cross-analysis of the results and combinatorial consideration with extra experimental information, there is no easy way to assess the confidence of the predicted CRMs. This limits the genome-wide understanding of CRMs. Description It is known that transcription factor binding and epigenetic profiles tend to determine functions of CRMs in gene transcriptional regulation. Thus integration of the genome-wide epigenetic profiles with systematically predicted CRMs can greatly help researchers evaluate and decipher the prediction confidence and possible transcriptional regulatory functions of these potential CRMs. However, these data are still fragmentary in the literatures. Here we performed the computational genome-wide screening for potential CRMs using different prediction tools and constructed the pioneer database, cisMEP (cis-regulatory module epigenetic profile database), to integrate these computationally identified CRMs with genomic epigenetic profile data. cisMEP collects the literature-curated TFBS location data and nine genres of epigenetic data for assessing the confidence of these potential CRMs and deciphering the possible CRM functionality. Conclusions cisMEP aims to provide a user-friendly interface for researchers to assess the confidence of different potential CRMs and to understand the functions of CRMs through experimentally-identified epigenetic profiles. The deposited potential CRMs and experimental epigenetic profiles for confidence assessment provide experimentally testable hypotheses for the molecular mechanisms of metazoan gene regulation. We believe that the information deposited in cisMEP will greatly facilitate the comparative usage of different CRM prediction tools and will help biologists to study the modular regulatory mechanisms between different TFs and their target genes.

MicroRNAs (miRNAs) are functional RNA molecules which play important roles in the post-transcriptional regulation. miRNAs regulate their target genes by repressing translation or inducing degradation of the target genes' mRNAs. Many databases have been constructed to provide computationally predicted miRNA targets. However, they cannot provide the miRNA targets expressed in a specific tissue and related to a specific disease at the same time. Moreover, they cannot provide the common targets of multiple miRNAs and the common miRNAs of multiple genes at the same time. To solve these two problems, we construct a database called CSmiRTar (Condition-Specific miRNA Targets). CSmiRTar collects computationally predicted targets of 2588 human miRNAs and 1945 mouse miRNAs from four most widely used miRNA target prediction databases (miRDB, TargetScan, microRNA.org and DIANA-microT) and implements functional filters which allows users to search (i) a miRNA's targets expressed in a specific tissue or/and related to a specific disease, (ii) multiple miRNAs' common targets expressed in a specific tissue or/and related to a specific disease, (iii) a gene's miRNAs related to a specific disease, and (iv) multiple genes' common miRNAs related to a specific disease. We believe that CSmiRTar will be a useful database for biologists to study the molecular mechanisms of post-transcriptional regulation in human or mouse. CSmiRTar is available at http://cosbi.ee.ncku.edu.tw/CSmiRTar/ or http://cosbi4.ee.ncku.edu.tw/CSmiRTar/.

Transcription factor binding site (TFBS) identification plays an important role in deciphering gene regulatory codes. With comprehensive knowledge of TFBSs, one can understand molecular mechanisms of gene regulation. In the recent decades, various computational approaches have been proposed to predict TFBSs in the genome. The TFBS dataset of a TF generated by each algorithm is a ranked list of predicted TFBSs of that TF, where top ranked TFBSs are statistically significant ones. However, whether these statistically significant TFBSs are functional (i.e. biologically relevant) is still unknown. Here we develop a post-processor, called the functional propensity calculator (FPC), to assign a functional propensity to each TFBS in the existing computationally predicted TFBS datasets. It is known that functional TFBSs reveal strong positional preference towards the transcriptional start site (TSS). This motivates us to take TFBS position relative to the TSS as the key idea in building our FPC. Based on our calculated functional propensities, the TFBSs of a TF in the original TFBS dataset could be reordered, where top ranked TFBSs are now the ones with high functional propensities. To validate the biological significance of our results, we perform three published statistical tests to assess the enrichment of Gene Ontology (GO) terms, the enrichment of physical protein-protein interactions, and the tendency of being co-expressed. The top ranked TFBSs in our reordered TFBS dataset outperform the top ranked TFBSs in the original TFBS dataset, justifying the effectiveness of our post-processor in extracting functional TFBSs from the original TFBS dataset. More importantly, assigning functional propensities to putative TFBSs enables biologists to easily identify which TFBSs in the promoter of interest are likely to be biologically relevant and are good candidates to do further detailed experimental investigation. The FPC is implemented as a web tool at http://santiago.ee.ncku.edu.tw/FPC/.

The advance of high-throughput experimental technologies generates many gene sets with different biological meanings, where many important insights can only be extracted by identifying the biological (regulatory/functional) features that are distinct between different gene sets (e.g. essential vs. non-essential genes, TATA box-containing vs. TATA box-less genes, induced vs. repressed genes under certain biological conditions). Although many servers have been developed to identify enriched features in a gene set, most of them were designed to analyze one gene set at a time but cannot compare two gene sets. Moreover, the features used in existing servers were mainly focused on functional annotations (GO terms), pathways, transcription factor binding sites (TFBSs) and/or protein-protein interactions (PPIs). In yeast, various important regulatory features, including promoter bendability, nucleosome occupancy, 5'-UTR length, and TF-gene regulation evidence, are available but have not been used in any enrichment analysis servers. This motivates us to develop the Yeast Genes Analyzer (YGA), a web server that simultaneously analyzes various biological (regulatory/functional) features of two gene sets and performs statistical tests to identify the distinct features between them. Many well-studied gene sets such as essential, stress-response, TATA box-containing and cell cycle genes were pre-compiled in YGA for users, if they have only one gene set, to compare with. In comparison with the existing enrichment analysis servers, YGA tests more comprehensive regulatory features (e.g. promoter bendability, nucleosome occupancy, 5'-UTR length, experimental evidence of TF-gene binding and TF-gene regulation) and functional features (e.g. PPI, GO terms, pathways and functional groups of genes, including essential/non-essential genes, stress-induced/-repressed genes, TATA box-containing/-less genes, occupied/depleted proximal-nucleosome genes and cell cycle genes). Furthermore, YGA uses various statistical tests to provide objective comparison measures. The two major contributions of YGA, comprehensive features and statistical comparison, help to mine important information that cannot be obtained from other servers. The sophisticated analysis tools of YGA can identify distinct biological features between two gene sets, which help biologists to form new hypotheses about the underlying biological mechanisms responsible for the observed difference between these two gene sets. YGA can be accessed from the following web pages: http://cosbi.ee.ncku.edu.tw/yga/ and http://yga.ee.ncku.edu.tw/.

In eukaryotic cells, transcriptional regulation of gene expression is usually achieved by cooperative transcription factors (TFs). Therefore, knowing cooperative TFs is the first step toward uncovering the molecular mechanisms of gene expression regulation. Many algorithms based on different rationales have been proposed to predict cooperative TF pairs in yeast. Although various types of rationales have been used in the existing algorithms, functional coherence is not yet used. This prompts us to develop a new algorithm based on functional coherence and similarity of the target gene sets to identify cooperative TF pairs in yeast. The proposed algorithm predicted 40 cooperative TF pairs. Among them, three (Pdc2-Thi2, Hot1-Msn1 and Leu3-Met28) are novel predictions, which have not been predicted by any existing algorithms. Strikingly, two (Pdc2-Thi2 and Hot1-Msn1) of the three novel predictions have been experimentally validated, demonstrating the power of the proposed algorithm. Moreover, we show that the predictions of the proposed algorithm are more biologically meaningful than the predictions of 17 existing algorithms under four evaluation indices. In summary, our study suggests that new algorithms based on novel rationales are worthy of developing for detecting previously unidentifiable cooperative TF pairs.

In eukaryotic cells, transcriptional regulation of gene expression is usually accomplished by cooperative Transcription Factors (TFs). Therefore, knowing cooperative TFs is helpful for uncovering the mechanisms of transcriptional regulation. In yeast, many cooperative TF pairs have been predicted by various algorithms in the literature. However, until now, there is still no database which collects the predicted yeast cooperative TFs from existing algorithms. This prompts us to construct Cooperative Transcription Factors Database (CoopTFD), which has a comprehensive collection of 2622 predicted cooperative TF pairs (PCTFPs) in yeast from 17 existing algorithms. For each PCTFP, our database also provides five types of validation information: (i) the algorithms which predict this PCTFP, (ii) the publications which experimentally show that this PCTFP has physical or genetic interactions, (iii) the publications which experimentally study the biological roles of both TFs of this PCTFP, (iv) the common Gene Ontology (GO) terms of this PCTFP and (v) the common target genes of this PCTFP. Based on the provided validation information, users can judge the biological plausibility of a PCTFP of interest. We believe that CoopTFD will be a valuable resource for yeast biologists to study the combinatorial regulation of gene expression controlled by cooperative TFs. Database URL: http://cosbi.ee.ncku.edu.tw/CoopTFD/ or http://cosbi2.ee.ncku.edu.tw/CoopTFD/

Background Transcription factor knockout microarrays (TFKMs) provide useful information about gene regulation. By using statistical methods for detecting differentially expressed genes between the gene expression microarray data of the mutant and wild type strains, the TF knockout targets of the knocked-out TF can be identified. However, the identified TF knockout targets may contain a certain amount of false positives due to the experimental noises inherent in the high-throughput microarray technology. Even if the identified TF knockout targets are true, the molecular mechanisms of how a TF regulates its TF knockout targets remain unknown by this kind of statistical approaches. Results To solve these two problems, we developed a method to filter out the false positives in the original TF knockout targets (identified by statistical approaches) so that the biologically interpretable TF knockout targets can be extracted. Our method can further generate experimentally testable hypotheses of the molecular mechanisms of how a TF regulates its biologically interpretable TF knockout targets. The details of our method are as follows. First, a TF binding network was constructed using the ChIP-chip data deposited in the YEASTRACT database. Then for each original TF knockout target, it is said to be biologically interpretable if a path (in the TF binding network) from the knocked-out TF to this target could be identified by our path search algorithm. The identified path explains how the TF may regulate this target either directly by binding to its promoter or indirectly through intermediate TFs. After checking all the original TF knockout targets, the biologically interpretable ones could be extracted and the false positives could be filtered out. We validated the biological significance of our refined (i.e., biologically interpretable) TF knockout targets by assessing their functional enrichment, expression coherence, and the prevalence of protein-protein interactions. Our refined TF knockout targets outperform the original TF knockout targets across all measures. Conclusions By jointly analyzing the TFKM and ChIP-chip data, our method can extract the biologically interpretable TF knockout targets by identifying paths (in the TF binding network) from the knocked-out TF to these targets. The identified paths form experimentally testable hypotheses regarding the molecular mechanisms of how a TF may regulate its knockout targets. About seven hundred hypotheses generated by our methods have been experimentally validated in the literature. Our work demonstrates that integrating different data sources is a powerful approach to study complex biological systems.

To understand the cell cycle process, it is important to identify the cell cycle transcription factors (TFs) that regulate the expression of cell cycle-regulated genes. We developed a method to identify cell cycle TFs in yeast by integrating current ChIP-chip, mutant, transcription factor binding site (TFBS), and cell cycle gene expression data. We identified 17 cell cycle TFs, 12 of which are known cell cycle TFs, while the remaining five (Ash1, Rlm1, Ste12, Stp1, Tec1) are putative novel cell cycle TFs. Several lines of evidence are provided to verify our predictions. Besides, a high confidence TF-gene regulatory matrix is derived as a byproduct of our method. Moreover, we show that our method performs better than four existing methods for identifying yeast cell cycle TFs. Finally, an application of our method to different cell cycle gene expression datasets suggests that our method is robust.

Background Eukaryotic cell cycle is a complex process and is precisely regulated at many levels. Many genes specific to the cell cycle are regulated transcriptionally and are expressed just before they are needed. To understand the cell cycle process, it is important to identify the cell cycle transcription factors (TFs) that regulate the expression of cell cycle-regulated genes. Results We developed a method to identify cell cycle TFs in yeast by integrating current ChIP-chip, mutant, transcription factor binding site (TFBS), and cell cycle gene expression data. We identified 17 cell cycle TFs, 12 of which are known cell cycle TFs, while the remaining five (Ash1, Rlm1, Ste12, Stp1, Tec1) are putative novel cell cycle TFs. For each cell cycle TF, we assigned specific cell cycle phases in which the TF functions and identified the time lag for the TF to exert regulatory effects on its target genes. We also identified 178 novel cell cycle-regulated genes, among which 59 have unknown functions, but they may now be annotated as cell cycle-regulated genes. Most of our predictions are supported by previous experimental or computational studies. Furthermore, a high confidence TF-gene regulatory matrix is derived as a byproduct of our method. Each TF-gene regulatory relationship in this matrix is supported by at least three data sources: gene expression, TFBS, and ChIP-chip or/and mutant data. We show that our method performs better than four existing methods for identifying yeast cell cycle TFs. Finally, an application of our method to different cell cycle gene expression datasets suggests that our method is robust. Conclusion Our method is effective for identifying yeast cell cycle TFs and cell cycle-regulated genes. Many of our predictions are validated by the literature. Our study shows that integrating multiple data sources is a powerful approach to studying complex biological systems.

Background ChIP-chip data, which indicate binding of transcription factors (TFs) to DNA regions in vivo, are widely used to reconstruct transcriptional regulatory networks. However, the binding of a TF to a gene does not necessarily imply regulation. Thus, it is important to develop methods to identify regulatory targets of TFs from ChIP-chip data. Results We developed a method, called Temporal Relationship Identification Algorithm (TRIA), which uses gene expression data to identify a TF's regulatory targets among its binding targets inferred from ChIP-chip data. We applied TRIA to yeast cell cycle microarray data and identified many plausible regulatory targets of cell cycle TFs. We validated our predictions by checking the enrichments for functional annotation and known cell cycle genes. Moreover, we showed that TRIA performs better than two published methods (MA-Network and MFA). It is known that co-regulated genes may not be co-expressed. TRIA has the ability to identify subsets of highly co-expressed genes among the regulatory targets of a TF. Different functional roles are found for different subsets, indicating the diverse functions a TF could have. Finally, for a control, we showed that TRIA also performs well for cell-cycle irrelevant TFs. Conclusion Finding the regulatory targets of TFs is important for understanding how cells change their transcription program to adapt to environmental stimuli. Our algorithm TRIA is helpful for achieving this purpose.

Background Biologists are puzzled by the extremely low percentage (3%) of the binding targets of a yeast transcription factor (TF) affected when the TF is knocked out, a phenomenon observed by comparing the TF binding dataset and TF knockout effect dataset. Results This study gives a plausible biological explanation of this counterintuitive phenomenon. Our analyses find that TFs with high functional redundancy show significantly lower percentage than do TFs with low functional redundancy. This suggests that functional redundancy may lead to one TF compensating for another, thus masking the TF knockout effect on the binding targets of the knocked-out TF. In addition, we show that seven classes of genes (lowly expressed genes, TATA box-less genes, genes containing a nucleosome-free region immediately upstream of the transcriptional start site (TSS), genes with low transcriptional plasticity, genes with a low number of bound TFs, genes with a low number of TFBSs, and genes with a short average distance of TFBSs to the TSS) are insensitive to the knockout of their promoter-binding TFs, providing clues for finding other biological explanations of the surprisingly low percentage of the binding targets of a TF affected when the TF is knocked out. Conclusions This study shows that one property of TFs (functional redundancy) and seven properties of genes (expression level, TATA box, nucleosome, transcriptional plasticity, the number of bound TFs, the number of TFBSs, and the average distance of TFBSs to the TSS) may be useful for explaining a counterintuitive phenomenon: most binding targets of a yeast transcription factor are not affected when the transcription factor is knocked out.

Background A transcriptional regulatory module (TRM) is a set of genes that is regulated by a common set of transcription factors (TFs). By organizing the genome into TRMs, a living cell can coordinate the activities of many genes and carry out complex functions. Therefore, identifying TRMs is helpful for understanding gene regulation. Results Integrating gene expression and ChIP-chip data, we develop a method, called MOdule Finding Algorithm (MOFA), for reconstructing TRMs of the yeast cell cycle. MOFA identified 87 TRMs, which together contain 336 distinct genes regulated by 40 TFs. Using various kinds of data, we validated the biological relevance of the identified TRMs. Our analysis shows that different combinations of a fairly small number of TFs are responsible for regulating a large number of genes involved in different cell cycle phases and that there may exist crosstalk between the cell cycle and other cellular processes. MOFA is capable of finding many novel TF-target gene relationships and can determine whether a TF is an activator or/and a repressor. Finally, MOFA refines some clusters proposed by previous studies and provides a better understanding of how the complex expression program of the cell cycle is regulated. Conclusion MOFA was developed to reconstruct TRMs of the yeast cell cycle. Many of these TRMs are in agreement with previous studies. Further, MOFA inferred many interesting modules and novel TF combinations. We believe that computational analysis of multiple types of data will be a powerful approach to studying complex biological systems when more and more genomic resources such as genome-wide protein activity data and protein-protein interaction data become available.

Regulatory targets of transcription factors (TFs) can be identified by the TF perturbation experiments, which reveal the expression changes owing to the perturbation (deletion or overexpression) of TFs. But the identified targets of a given TF consist of both direct and indirect regulatory targets. It has been shown that most of the TFPE-identified regulatory targets are indirect, indicating that TF-gene regulation is mainly through transcriptional regulatory pathways (TRPs) consisting of intermediate TFs. Without identification of these TRPs, it is not easy to understand how a TF regulates its indirect targets. Because there is no such database depositing the potential TRPs for Saccharomyces cerevisiae now, this motivates us to construct the YTRP (Yeast Transcriptional Regulatory Pathway) database. For each TF-gene regulatory pair under different experimental conditions, all possible TRPs in two underlying networks (constructed using experimentally verified TF-gene binding pairs and TF-gene regulatory pairs from the literature) for the specified experimental conditions were automatically enumerated by TRP mining procedures developed from the graph theory. The enumerated TRPs of a TF-gene regulatory pair provide experimentally testable hypotheses for the molecular mechanisms behind a TF and its regulatory target. YTRP is available online at http://cosbi3.ee.ncku.edu.tw/YTRP/. We believe that the TRPs deposited in this database will greatly improve the usefulness of TFPE data for yeast biologists to study the regulatory mechanisms between a TF and its knocked-out targets. Database URL: http://cosbi3.ee.ncku.edu.tw/YTRP/

Background Investigating association between genes can be used in understanding the relations of genes in biological processes. STRING and GeneMANIA are two well-known web tools which can provide a list of associated genes of a query gene based on diverse biological associations such as co-expression, co-localization, co-citation and so on. However, the transcriptional regulation association and mutant phenotype association have not been used in these two web tools. Since the comprehensive transcription factor (TF)-gene binding data, TF-gene regulation data and mutant phenotype data are available in yeast, we developed a web tool called YAGM (Yeast Associated Genes Miner) which constructed the transcriptional regulation association, mutant phenotype association and five commonly used biological associations to mine a list of associated genes of a query yeast gene. Description In YAGM, we collected seven kinds of datasets including TF-gene binding (TFB) data, TF-gene regulation (TFR) data, mutant phenotype (MP) data, functional annotation (FA) data, physical interaction (PI) data, genetic interaction (GI) data, and literature evidence (LE) data. Then by using the hypergeometric test to calculate the association scores of all gene pairs in yeast, we constructed seven biological associations including two transcriptional regulation associations (TFB association and TFR association), MP association, FA association, PI association, GI association, and LE association. Moreover, the expression profile association from SPELL database was also included in YAGM. When using YAGM, users can input a query gene and choose any possible subsets of the eight biological associations, then a list of associated genes of the query gene will be returned based on the chosen biological associations. Conclusions In this study, we presented the YAGM which provides eight biological associations for mining associated genes of a query gene in yeast. Among the eight biological associations constructed in YAGM, three (TFB association, TFR association, and MP association) are novel ones. By comparing the query results of two well-known web tools (STRING and GeneMANIA), we found that YAGM can find out distinct associated genes of a query gene. That is, YAGM can provide alternative candidates of associated genes for biologists to do further experimental investigation. We believe that YAGM will be a useful web tool for yeast biologists. YAGM is available online at http://cosbi3.ee.ncku.edu.tw/yagm/.

Background Transcriptional regulation of gene expression in eukaryotes is usually accomplished by cooperative transcription factors (TFs). Computational identification of cooperative TF pairs has become a hot research topic and many algorithms have been proposed in the literature. A typical algorithm for predicting cooperative TF pairs has two steps. (Step 1) Define the targets of each TF under study. (Step 2) Design a measure for calculating the cooperativity of a TF pair based on the targets of these two TFs. While different algorithms have distinct sophisticated cooperativity measures, the targets of a TF are usually defined using ChIP-chip data. However, there is an inherent weakness in using ChIP-chip data to define the targets of a TF. ChIP-chip analysis can only identify the binding targets of a TF but it cannot distinguish the true regulatory from the binding but non-regulatory targets of a TF. Results This work is the first study which aims to investigate whether the performance of computational identification of cooperative TF pairs could be improved by using a more biologically relevant way to define the targets of a TF. For this purpose, we propose four simple algorithms, all of which consist of two steps. (Step 1) Define the targets of a TF using (i) ChIP-chip data in the first algorithm, (ii) TF binding data in the second algorithm, (iii) TF perturbation data in the third algorithm, and (iv) the intersection of TF binding and TF perturbation data in the fourth algorithm. Compared with the first three algorithms, the fourth algorithm uses a more biologically relevant way to define the targets of a TF. (Step 2) Measure the cooperativity of a TF pair by the statistical significance of the overlap of the targets of these two TFs using the hypergeometric test. By adopting four existing performance indices, we show that the fourth proposed algorithm (PA4) significantly out performs the other three proposed algorithms. This suggests that the computational identification of cooperative TF pairs is indeed improved when using a more biologically relevant way to define the targets of a TF. Strikingly, the prediction results of our simple PA4 are more biologically meaningful than those of the 12 existing sophisticated algorithms in the literature, all of which used ChIP-chip data to define the targets of a TF. This suggests that properly defining the targets of a TF may be more important than designing sophisticated cooperativity measures. In addition, our PA4 has the power to predict several experimentally validated cooperative TF pairs, which have not been successfully predicted by any existing algorithms in the literature. Conclusions This study shows that the performance of computational identification of cooperative TF pairs could be improved by using a more biologically relevant way to define the targets of a TF. The main contribution of this study is not to propose another new algorithm but to provide a new thinking for the research of computational identification of cooperative TF pairs. Researchers should put more effort on properly defining the targets of a TF (i.e. Step 1) rather than totally focus on designing sophisticated cooperativity measures (i.e. Step 2). The lists of TF target genes, the Matlab codes and the prediction results of the four proposed algorithms could be downloaded from our companion website http://cosbi3.ee.ncku.edu.tw/TFI/

Background Comprehensive characterization of the phosphoproteome in living cells is critical in signal transduction research. But the low abundance of phosphopeptides among the total proteome in cells remains an obstacle in mass spectrometry-based proteomic analysis. To provide a solution, an alternative analytic strategy to confidently identify phosphorylated peptides by using the alkaline phosphatase (AP) treatment combined with high-resolution mass spectrometry was provided. While the process is applicable, the key integration along the pipeline was mostly done by tedious manual work. Results We developed a software toolkit, iPhos, to facilitate and streamline the work-flow of AP-assisted phosphoproteome characterization. The iPhos tookit includes one assister and three modules. The iPhos Peak Extraction Assister automates the batch mode peak extraction for multiple liquid chromatography mass spectrometry (LC-MS) runs. iPhos Module-1 can process the peak lists extracted from the LC-MS analyses derived from the original and dephosphorylated samples to mine out potential phosphorylated peptide signals based on mass shift caused by the loss of some multiples of phosphate groups. And iPhos Module-2 provides customized inclusion lists with peak retention time windows for subsequent targeted LC-MS/MS experiments. Finally, iPhos Module-3 facilitates to link the peptide identifications from protein search engines to the quantification results from pattern-based label-free quantification tools. We further demonstrated the utility of the iPhos toolkit on the data of human metastatic lung cancer cells (CL1-5). Conclusions In the comparison study of the control group of CL1-5 cell lysates and the treatment group of dasatinib-treated CL1-5 cell lysates, we demonstrated the applicability of the iPhos toolkit and reported the experimental results based on the iPhos-facilitated phosphoproteome investigation. And further, we also compared the strategy with pure DDA-based LC-MS/MS phosphoproteome investigation. The results of iPhos-facilitated targeted LC-MS/MS analysis convey more thorough and confident phosphopeptide identification than the results of pure DDA-based analysis.

Background Transcriptional regulation of gene expression is usually accomplished by multiple interactive transcription factors (TFs). Therefore, it is crucial to understand the precise cooperative interactions among TFs. Various kinds of experimental data including ChIP-chip, TF binding site (TFBS), gene expression, TF knockout and protein-protein interaction data have been used to identify cooperative TF pairs in existing methods. The nucleosome occupancy data is not yet used for this research topic despite that several researches have revealed the association between nucleosomes and TFBSs. Results In this study, we developed a novel method to infer the cooperativity between two TFs by integrating the TF-gene documented regulation, TFBS and nucleosome occupancy data. TF-gene documented regulation and TFBS data were used to determine the target genes of a TF, and the genome-wide nucleosome occupancy data was used to assess the nucleosome occupancy on TFBSs. Our method identifies cooperative TF pairs based on two biologically plausible assumptions. If two TFs cooperate, then (i) they should have a significantly higher number of common target genes than random expectation and (ii) their binding sites (in the promoters of their common target genes) should tend to be co-depleted of nucleosomes in order to make these binding sites simultaneously accessible to TF binding. Each TF pair is given a cooperativity score by our method. The higher the score is, the more likely a TF pair has cooperativity. Finally, a list of 27 cooperative TF pairs has been predicted by our method. Among these 27 TF pairs, 19 pairs are also predicted by existing methods. The other 8 pairs are novel cooperative TF pairs predicted by our method. The biological relevance of these 8 novel cooperative TF pairs is justified by the existence of protein-protein interactions and co-annotation in the same MIPS functional categories. Moreover, we adopted three performance indices to compare our predictions with 11 existing methods' predictions. We show that our method performs better than these 11 existing methods in identifying cooperative TF pairs in yeast. Finally, the cooperative TF network constructed from the 27 predicted cooperative TF pairs shows that our method has the power to find cooperative TF pairs of different biological processes. Conclusion Our method is effective in identifying cooperative TF pairs in yeast. Many of our predictions are validated by the literature, and our method outperforms 11 existing methods. We believe that our study will help biologists to understand the mechanisms of transcriptional regulation in eukaryotic cells.

Microarray data are usually peppered with missing values due to various reasons. However, most of the downstream analyses for microarray data require complete datasets. Therefore, accurate algorithms for missing value estimation are needed for improving the performance of microarray data analyses. Although many algorithms have been developed, there are many debates on the selection of the optimal algorithm. The studies about the performance comparison of different algorithms are still incomprehensive, especially in the number of benchmark datasets used, the number of algorithms compared, the rounds of simulation conducted, and the performance measures used. In this paper, we performed a comprehensive comparison by using (I) thirteen datasets, (II) nine algorithms, (III) 110 independent runs of simulation, and (IV) three types of measures to evaluate the performance of each imputation algorithm fairly. First, the effects of different types of microarray datasets on the performance of each imputation algorithm were evaluated. Second, we discussed whether the datasets from different species have different impact on the performance of different algorithms. To assess the performance of each algorithm fairly, all evaluations were performed using three types of measures. Our results indicate that the performance of an imputation algorithm mainly depends on the type of a dataset but not on the species where the samples come from. In addition to the statistical measure, two other measures with biological meanings are useful to reflect the impact of missing value imputation on the downstream data analyses. Our study suggests that local-least-squares-based methods are good choices to handle missing values for most of the microarray datasets. In this work, we carried out a comprehensive comparison of the algorithms for microarray missing value imputation. Based on such a comprehensive comparison, researchers could choose the optimal algorithm for their datasets easily. Moreover, new imputation algorithms could be compared with the existing algorithms using this comparison strategy as a standard protocol. In addition, to assist researchers in dealing with missing values easily, we built a web-based and easy-to-use imputation tool, MissVIA ( http://cosbi.ee.ncku.edu.tw/MissVIA ), which supports many imputation algorithms. Once users upload a real microarray dataset and choose the imputation algorithms, MissVIA will determine the optimal algorithm for the users' data through a series of simulations, and then the imputed results can be downloaded for the downstream data analyses.

Background Missing values commonly occur in the microarray data, which usually contain more than 5% missing values with up to 90% of genes affected. Inaccurate missing value estimation results in reducing the power of downstream microarray data analyses. Many types of methods have been developed to estimate missing values. Among them, the regression-based methods are very popular and have been shown to perform better than the other types of methods in many testing microarray datasets. Results To further improve the performances of the regression-based methods, we propose shrinkage regression-based methods. Our methods take the advantage of the correlation structure in the microarray data and select similar genes for the target gene by Pearson correlation coefficients. Besides, our methods incorporate the least squares principle, utilize a shrinkage estimation approach to adjust the coefficients of the regression model, and then use the new coefficients to estimate missing values. Simulation results show that the proposed methods provide more accurate missing value estimation in six testing microarray datasets than the existing regression-based methods do. Conclusions Imputation of missing values is a very important aspect of microarray data analyses because most of the downstream analyses require a complete dataset. Therefore, exploring accurate and efficient methods for estimating missing values has become an essential issue. Since our proposed shrinkage regression-based methods can provide accurate missing value estimation, they are competitive alternatives to the existing regression-based methods.