Time-series Expression Profiles Research Articles

MirTime: identifying condition-specific targets of microRNA in time-series transcript data using Gaussian process model and spherical vector clustering.

MicroRNAs, small noncoding RNAs, are conserved in many species, and they are key regulators that mediate post-transcriptional gene silencing. Since biologists cannot perform experiments for each of target genes of thousands of microRNAs in numerous specific conditions, prediction on microRNA target genes has been extensively investigated. A general framework is a two-step process of selecting target candidates based on sequence and binding energy features and then predicting targets based on negative correlation of microRNAs and their targets. However, there are few methods that are designed for target predictions using time-series gene expression data. In this article, we propose a new pipeline, mirTime, that predicts microRNA targets by integrating sequence features and time-series expression profiles in a specific experimental condition. The most important feature of mirTime is that it uses the Gaussian process regression model to measure data at unobserved or unpaired time points. In experiments with two datasets in different experimental conditions and cell types, condition-specific target modules reported in the original papers were successfully predicted with our pipeline. The context specificity of target modules was assessed with three (correlation-based, target gene-based and network-based) evaluation criteria. mirTime showed better performance than existing expression-based microRNA target prediction methods in all three criteria. mirTime is available at https://github.com/mirTime/mirtime. Supplementary data are available at Bioinformatics online.

Automated transition analysis of activated gene regulation during diauxic nutrient shift in Escherichia coli and adipocyte differentiation in mouse cells

BackgroundComprehensively understanding the dynamics of biological systems is among the biggest current challenges in biology and medicine. To acquire this understanding, researchers have measured the time-series expression profiles of cell lines of various organisms. Biological technologies have also drastically improved, providing a huge amount of information with support from bioinformatics and systems biology. However, the transitions between the activation and inactivation of gene regulations, at the temporal resolution of single time points, are difficult to extract from time-course gene expression profiles.ResultsOur proposed method reports the activation period of each gene regulation from gene expression profiles and a gene regulatory network. The correctness and effectiveness of the method were validated by analyzing the diauxic shift from glucose to lactose in Escherichia coli. The method completely detected the three periods of the shift; 1) consumption of glucose as nutrient source, 2) the period of seeking another nutrient source and 3) consumption of lactose as nutrient source. We then applied the method to mouse adipocyte differentiation data. Cell differentiation into adipocytes is known to involve two waves of the gene regulation cascade, and sub-waves are predicted. From the gene expression profiles of the cell differentiation process from ES to adipose cells (62 time points), our method acquired four periods; three periods covering the two known waves of the cascade, and a final period of gene regulations when the differentiation to adipocytes was completed.ConclusionsOur proposed method identifies the transitions of gene regulations from time-series gene expression profiles. Dynamic analyses are essential for deep understanding of biological systems and for identifying the causes of the onset of diseases such as diabetes and osteoporosis. The proposed method can greatly contribute to the progress of biology and medicine.

Clustering gene expression time series data using an infinite Gaussian process mixture model.

Transcriptome-wide time series expression profiling is used to characterize the cellular response to environmental perturbations. The first step to analyzing transcriptional response data is often to cluster genes with similar responses. Here, we present a nonparametric model-based method, Dirichlet process Gaussian process mixture model (DPGP), which jointly models data clusters with a Dirichlet process and temporal dependencies with Gaussian processes. We demonstrate the accuracy of DPGP in comparison to state-of-the-art approaches using hundreds of simulated data sets. To further test our method, we apply DPGP to published microarray data from a microbial model organism exposed to stress and to novel RNA-seq data from a human cell line exposed to the glucocorticoid dexamethasone. We validate our clusters by examining local transcription factor binding and histone modifications. Our results demonstrate that jointly modeling cluster number and temporal dependencies can reveal shared regulatory mechanisms. DPGP software is freely available online at https://github.com/PrincetonUniversity/DP_GP_cluster.

Hilbert-Schmidt and Sobol sensitivity indices for static and time series Wnt signaling measurements in colorectal cancer - part A

BackgroundEver since the accidental discovery of Wingless [Sharma R.P., Drosophila information service, 1973, 50, p 134], research in the field of Wnt signaling pathway has taken significant strides in wet lab experiments and various cancer clinical trials, augmented by recent developments in advanced computational modeling of the pathway. Information rich gene expression profiles reveal various aspects of the signaling pathway and help in studying different issues simultaneously. Hitherto, not many computational studies exist which incorporate the simultaneous study of these issues.ResultsThis manuscript ∙ explores the strength of contributing factors in the signaling pathway, ∙ analyzes the existing causal relations among the inter/extracellular factors effecting the pathway based on prior biological knowledge and ∙ investigates the deviations in fold changes in the recently found prevalence of psychophysical laws working in the pathway. To achieve this goal, local and global sensitivity analysis is conducted on the (non)linear responses between the factors obtained from static and time series expression profiles using the density (Hilbert-Schmidt Information Criterion) and variance (Sobol) based sensitivity indices.ConclusionThe results show the advantage of using density based indices over variance based indices mainly due to the former’s employment of distance measures & the kernel trick via Reproducing kernel Hilbert space (RKHS) that capture nonlinear relations among various intra/extracellular factors of the pathway in a higher dimensional space. In time series data, using these indices it is now possible to observe where in time, which factors get influenced & contribute to the pathway, as changes in concentration of the other factors are made. This synergy of prior biological knowledge, sensitivity analysis & representations in higher dimensional spaces can facilitate in time based administration of target therapeutic drugs & reveal hidden biological information within colorectal cancer samples.

Dynamic transcriptomic analysis in hircine longissimus dorsi muscle from fetal to neonatal development stages.

Muscle growth and development from fetal to neonatal stages consist of a series of delicately regulated and orchestrated changes in expression of genes. In this study, we performed whole transcriptome profiling based on RNA-Seq of caprine longissimus dorsi muscle tissue obtained from prenatal stages (days 45, 60, and 105 of gestation) and neonatal stage (the 3-day-old newborn) to identify genes that are differentially expressed and investigate their temporal expression profiles. A total of 3276 differentially expressed genes (DEGs) were identified (Q value <0.01). Time-series expression profile clustering analysis indicated that DEGs were significantly clustered into eight clusters which can be divided into two classes (Q value <0.05), class I profiles with downregulated patterns and class II profiles with upregulated patterns. Based on cluster analysis, GO enrichment analysis found that 75, 25, and 8 terms to be significantly enriched in biological process (BP), cellular component (CC), and molecular function (MF) categories in class I profiles, while 35, 21, and 8 terms to be significantly enriched in BP, CC, and MF in class II profiles. KEGG pathway analysis revealed that DEGs from class I profiles were significantly enriched in 22 pathways and the most enriched pathway was Rap1 signaling pathway. DEGs from class II profiles were significantly enriched in 17 pathways and the mainly enriched pathway was AMPK signaling pathway. Finally, six selected DEGs from our sequencing results were confirmed by qPCR. Our study provides a comprehensive understanding of the molecular mechanisms during goat skeletal muscle development from fetal to neonatal stages and valuable information for future studies of muscle development in goats.

Content-Based Search on Time-Series Microarray Databases Using Cluster-Based Fingerprints

Background: The rapid growth of gene expression databases has created a need for contentbased searches as an alternative to unstructured database queries using keyword- or metadata-based searches. Content-based searching is the ability to retrieve all experiments with similar gene expression patterns in a database regardless of the biological annotations provided for these experiments. Objective: While this concept is still in its infancy in a general context, in this study we focus on applying it to a specific subset of gene expression datasets, by only querying experiments involving time-series expression profiles. Method: To this end, we propose a novel experiment fingerprinting scheme obtained by clustering expression profiles, for content-based searching of time-series microarray experiments. To determine the retrieval ability of the proposed scheme, we performed a simulated information retrieval task on a large set of microarray experiments gathered from a public repository. The relevance between any two experiments was then defined using their commonalities based on annotated disease associations. Results and Conclusion: The results showed that relevant experiments can be more successfully retrieved using this new method compared with traditional differential expression-based methods. Keywords: Gene expression database, time-course data, time-series profile, content-based search, information retrieval, modelbased clustering.

Time-series expression profile analysis of fracture healing in young and old mice.

Bone fracture healing is a complex process, which is associated with several factors, including age and osteoporosis. Certain genes and biological processes that may contribute to fracture healing, have been identified following developments in systems biology and molecular biology technologies, which may benefit the treatment of bone fractures. The present study identified key genes, which may be important in fracture healing through bioinformatics analysis of gene microarray datasets from the Gene Expression Omnibus. Gene clusters, which were consistently up/downregulated through time following a fracture in young (6-week-old) mice and old (8-month-old retired breeders) mice were obtained via soft clustering of differentially expressed genes (DEGs) between samples at 1 and 3 days, 1 and 5 days, and 3 and 5 days post-fracture in the two age groups, based on the Mfuzz package of R. Functional enrichment analysis of gene clusters using the Database for Annotation, Visualization and Integrated Discovery indicated that biological processes and pathways, including those associated with bone development, skeletal system development, amino sugar and nucleotide sugar metabolism, were significantly enriched in these up/downregulated genes. Of note, a total of 207 overlapped consistently upregulated genes were obtained between the two age groups, whereas no overlap was identified between the two lists of consistently downregulated genes. The overlapped genes were found to be associated with the biological processes of blood vessel development, vasculature development and skeletal system development, compared with all genes in the clusters. In addition, certain genes, including epidermal growth factor-like domain multiple 6 (EGFL6), kazal-type serine peptidase inhibitor domain 1 (KAZALD1), olfactomedin 2B (OLFM2B), collagen type III α1 (COL3A1), collagen type II α1 (COL2A1), von Willebrand factor A domain-containing 1 (VWA1), elastin microfibril interfacer 1 (EMILIN1) and aggrecan (ACAN), of the extracellular matrix organization, a process performed at the cellular level and resulting in the assembly and arrangement of constituent parts, were confirmed to be associated with fracture healing via reverse transcription-quantitative polymerase chain reaction analysis. The present study identified certain genes and biological processes/pathways, which may be associated with fracture healing and may assist in fundamental investigations and treatment in the future.

Transcriptome analysis provides insights into wood formation during larch tree aging

Tree age affects wood formation and yield. However, the underlying mechanisms are poorly understood, particularly at the molecular level. In this study, we investigated the transcriptomic changes of the uppermost main stems of Larix kaempferi in an entire rotation period using the RNA-Seq method. In total, ∼151 million reads were obtained from the stems of 1-, 2-, 5-, 10-, 25-, and 50-year-old L. kaempferi trees. Combining these with the published Illumina sequencing reads, 299,637 assembled transcripts were generated, of which 161,232 were annotated. Time series expression profiling identified 12,927 transcripts as differentially expressed genes (DEGs); function enrichment analysis of these DEGs showed that 459 gene ontology terms in the biological process category were enriched. These terms were associated with the processes of wood formation, such as cell differentiation, growth and death, and its hormonal regulation. Based on the expression patterns of L. kaempferi homologues of genes associated with ethylene, calcium, and cell wall expansion and synthesis, the regulatory network of tracheid growth was outlined. Altogether, the comparative transcriptomic analysis reported here demonstrated the molecular aspects of aging effects on L. kaempferi wood formation. The identification of genes involved in the regulatory network of tracheid growth provides a means of investigating the regulation of wood formation in gymnosperm trees and also offers potential targets for genetic manipulation to improve the properties of xylem fibers.

Differential Gene Expression (DEX) and Alternative Splicing Events (ASE) for Temporal Dynamic Processes Using HMMs and Hierarchical Bayesian Modeling Approaches.

In gene expression profile, data analysis pipeline is categorized into four levels, major downstream tasks, i.e., (1) identification of differential expression; (2) clustering co-expression patterns; (3) classification of subtypes of samples; and (4) detection of genetic regulatory networks, are performed posterior to preprocessing procedure such as normalization techniques. To be more specific, temporal dynamic gene expression data has its inherent feature, namely, two neighboring time points (previous and current state) are highly correlated with each other, compared to static expression data which samples are assumed as independent individuals. In this chapter, we demonstrate how HMMs and hierarchical Bayesian modeling methods capture the horizontal time dependency structures in time series expression profiles by focusing on the identification of differential expression. In addition, those differential expression genes and transcript variant isoforms over time detected in core prerequisite steps can be generally further applied in detection of genetic regulatory networks to comprehensively uncover dynamic repertoires in the aspects of system biology as the coupled framework.

A Dynamic Multiagent Genetic Algorithm for Gene Regulatory Network Reconstruction Based on Fuzzy Cognitive Maps

In order to reconstruct large-scale gene regulatory networks (GRNs) with high accuracy, a robust evolutionary algorithm, a dynamic multiagent genetic algorithm (dMAGA), is proposed to reconstruct GRNs from time-series expression profiles based on fuzzy cognitive maps (FCMs) in this paper. The algorithm is labeled as dMAGAFCM-GRN. In dMAGAFCM-GRN, agents and their behaviors are designed with the intrinsic properties of GRN reconstruction problems in mind. All agents live in a lattice-like environment, and the neighbors of each agent are changed dynamically according to their energy in each generation. dMAGAFCM-GRN can learn continuous states directly for FCMs from data. In the experiments, the performance of dMAGAFCM-GRN is validated on both large-scale synthetic data and the benchmark DREAM3 and DREAM4. The experimental results show that dMAGAFCM-GRN is able to effectively learn FCMs with 200 nodes; that is, 40 000 weights need to be optimized. The systematic comparison with five existing algorithms shows that dMAGAFCM-GRN outperforms all other algorithms and can approximate the time series with high accuracy.

Reconstructing gene regulatory networks with a memetic-neural hybrid based on fuzzy cognitive maps

Reconstructing gene regulatory networks (GRNs) plays an important role in identifying the complicated regulatory relationships, uncovering regulatory patterns in cells, and gaining a systematic view for biological processes. In order to reconstruct large-scale GRNs accurately, in this paper, we first use fuzzy cognitive maps (FCMs), which are a kind of cognition fuzzy influence graphs based on fuzzy logic and neural networks, to model GRNs. Then, a novel hybrid method is proposed to reconstruct GRNs from time series expression profiles using memetic algorithm (MA) combined with neural network (NN), which is labeled as MANNFCM-GRN. In MANNFCM-GRN, the MA is used to determine regulatory connections in GRNs and the NN is used to determine the interaction strength of the regulatory connections. In the experiments, the performance of MANNFCM-GRN is validated on both synthetic data and the benchmark dataset DREAM3 and DREAM4. The experimental results demonstrate the efficacy of MANNFCM-GRN and show that MANNFCM-GRN can reconstruct GRNs with high accuracy without expert knowledge. The comparison with existing algorithms also shows that MANNFCM-GRN outperforms ant colony optimization, non-linear Hebbian learning, and real-coded genetic algorithms.

Comparative analysis of molecular mechanism of spinal cord injury with time based on bioinformatics data.

This study was designed to explore the molecular mechanisms of spinal cord injury (SCI) with time. The gene expression profile (GSE45006) including four non-injured spinal cord samples as sham-control group and 20 thoracic transected spinal cords samples as experimental group at different times was downloaded from Gene Expression Omnibus database. The time-course changes of the SCI-related differentially expressed genes (DEGs) were identified. In addition, time-series expression profile clusters of DEGs were obtained, followed by gene ontology (GO) and pathway enrichment analysis of the DEGs. Moreover, the transcriptional regulatory network was constructed. There were 1420, 492, 743, 568 and 533 DEGs respectively at d1, d3, w1, w2 and w8 compared with that of sham group. Importantly, 101 overlapped regulated DEGs were identified at five time points and 370 collaboratively regulated genes were identified in cluster 6. Significant functions of overlapped regulated DEGs were obtained including response to wounding and developmental process. In addition, the DEGs, such as CD14 molecule (CD14) and chemokine (C-C motif) ligand 2 (CCL2), were enriched mostly in the pathways related to tuberculosis, phagosome and NF-kappa B signaling pathway. From the transcriptional regulatory network, we identified some transription factors (TFs), including member of E26 transformation-specific (ETS) oncogene family (ELK1) and zinc finger and BTB domain containing 7A (Zbtb7a). The DEGs related to immune response during SCI may provide underlying targets for treatment of SCI. Moreover, the TFs ZBTB7A and ELK1 and their target gene (dual specificity phosphatase 18 (DUSP18)) might be therapeutic targets for the treatment of SCI.

Identification of Gene Networks with Time Delayed Regulation Based on Temporal Expression Profiles.

There are fundamental limitations in inferring the functional interaction structure of a gene (regulatory) network only from sequence information such as binding motifs. To overcome such limitations, various approaches have been developed to infer the functional interaction structure from expression profiles. However, most of them have not been so successful due to the experimental limitations and computational complexity. Hence, there is a pressing need to develop a simple but effective methodology that can systematically identify the functional interaction structure of a gene network from time-series expression profiles. In particular, we need to take into account the different time delay effects in gene regulation since they are ubiquitously present. We have considered a new experiment that measures the overall expression changes after a perturbation on a specific gene. Based on this experiment, we have proposed a new inference method that can take account of the time delay induced while the perturbation affects its primary target genes. Specifically, we have developed an algebraic equation from which we can identify the subnetwork structure around the perturbed gene. We have also analyzed the influence of time delay on the inferred network structure. The proposed method is particularly useful for identification of a gene network with small variations in the time delay of gene regulation.

Systems perspectives on erythromycin biosynthesis by comparative genomic and transcriptomic analyses of S. erythraea E3 and NRRL23338 strains

BackgroundS. erythraea is a Gram-positive filamentous bacterium used for the industrial-scale production of erythromycin A which is of high clinical importance. In this work, we sequenced the whole genome of a high-producing strain (E3) obtained by random mutagenesis and screening from the wild-type strain NRRL23338, and examined time-series expression profiles of both E3 and NRRL23338. Based on the genomic data and transcriptpmic data of these two strains, we carried out comparative analysis of high-producing strain and wild-type strain at both the genomic level and the transcriptomic level.ResultsWe observed a large number of genetic variants including 60 insertions, 46 deletions and 584 single nucleotide variations (SNV) in E3 in comparison with NRRL23338, and the analysis of time series transcriptomic data indicated that the genes involved in erythromycin biosynthesis and feeder pathways were significantly up-regulated during the 60 hours time-course. According to our data, BldD, a previously identified ery cluster regulator, did not show any positive correlations with the expression of ery cluster, suggesting the existence of alternative regulation mechanisms of erythromycin synthesis in S. erythraea. Several potential regulators were then proposed by integration analysis of genomic and transcriptomic data.ConclusionThis is a demonstration of the functional comparative genomics between an industrial S. erythraea strain and the wild-type strain. These findings help to understand the global regulation mechanisms of erythromycin biosynthesis in S. erythraea, providing useful clues for genetic and metabolic engineering in the future.

Inference of gene regulatory networks from time series by Tsallis entropy

BackgroundThe inference of gene regulatory networks (GRNs) from large-scale expression profiles is one of the most challenging problems of Systems Biology nowadays. Many techniques and models have been proposed for this task. However, it is not generally possible to recover the original topology with great accuracy, mainly due to the short time series data in face of the high complexity of the networks and the intrinsic noise of the expression measurements. In order to improve the accuracy of GRNs inference methods based on entropy (mutual information), a new criterion function is here proposed.ResultsIn this paper we introduce the use of generalized entropy proposed by Tsallis, for the inference of GRNs from time series expression profiles. The inference process is based on a feature selection approach and the conditional entropy is applied as criterion function. In order to assess the proposed methodology, the algorithm is applied to recover the network topology from temporal expressions generated by an artificial gene network (AGN) model as well as from the DREAM challenge. The adopted AGN is based on theoretical models of complex networks and its gene transference function is obtained from random drawing on the set of possible Boolean functions, thus creating its dynamics. On the other hand, DREAM time series data presents variation of network size and its topologies are based on real networks. The dynamics are generated by continuous differential equations with noise and perturbation. By adopting both data sources, it is possible to estimate the average quality of the inference with respect to different network topologies, transfer functions and network sizes.ConclusionsA remarkable improvement of accuracy was observed in the experimental results by reducing the number of false connections in the inferred topology by the non-Shannon entropy. The obtained best free parameter of the Tsallis entropy was on average in the range 2.5 ≤ q ≤ 3.5 (hence, subextensive entropy), which opens new perspectives for GRNs inference methods based on information theory and for investigation of the nonextensivity of such networks. The inference algorithm and criterion function proposed here were implemented and included in the DimReduction software, which is freely available at http://sourceforge.net/projects/dimreduction and http://code.google.com/p/dimreduction/.

Efficient parameter search for qualitative models of regulatory networks using symbolic model checking

Motivation: Investigating the relation between the structure and behavior of complex biological networks often involves posing the question if the hypothesized structure of a regulatory network is consistent with the observed behavior, or if a proposed structure can generate a desired behavior.Results: The above questions can be cast into a parameter search problem for qualitative models of regulatory networks. We develop a method based on symbolic model checking that avoids enumerating all possible parametrizations, and show that this method performs well on real biological problems, using the IRMA synthetic network and benchmark datasets. We test the consistency between IRMA and time-series expression profiles, and search for parameter modifications that would make the external control of the system behavior more robust.Availability: GNA and the IRMA model are available at http://ibis.inrialpes.fr/Contact: gregory.batt@inria.frSupplementary information: Supplementary data are available at Bioinformatics online.

Linear time-varying models can reveal non-linear interactions of biomolecular regulatory networks using multiple time-series data.

Inherent non-linearities in biomolecular interactions make the identification of network interactions difficult. One of the principal problems is that all methods based on the use of linear time-invariant models will have fundamental limitations in their capability to infer certain non-linear network interactions. Another difficulty is the multiplicity of possible solutions, since, for a given dataset, there may be many different possible networks which generate the same time-series expression profiles. A novel algorithm for the inference of biomolecular interaction networks from temporal expression data is presented. Linear time-varying models, which can represent a much wider class of time-series data than linear time-invariant models, are employed in the algorithm. From time-series expression profiles, the model parameters are identified by solving a non-linear optimization problem. In order to systematically reduce the set of possible solutions for the optimization problem, a filtering process is performed using a phase-portrait analysis with random numerical perturbations. The proposed approach has the advantages of not requiring the system to be in a stable steady state, of using time-series profiles which have been generated by a single experiment, and of allowing non-linear network interactions to be identified. The ability of the proposed algorithm to correctly infer network interactions is illustrated by its application to three examples: a non-linear model for cAMP oscillations in Dictyostelium discoideum, the cell-cycle data for Saccharomyces cerevisiae and a large-scale non-linear model of a group of synchronized Dictyostelium cells. The software used in this article is available from http://sbie.kaist.ac.kr/software

TWO-PASS IMPUTATION ALGORITHM FOR MISSING VALUE ESTIMATION IN GENE EXPRESSION TIME SERIES

Gene expression microarray experiments frequently generate datasets with multiple values missing. However, most of the analysis, mining, and classification methods for gene expression data require a complete matrix of gene array values. Therefore, the accurate estimation of missing values in such datasets has been recognized as an important issue, and several imputation algorithms have already been proposed to the biological community. Most of these approaches, however, are not particularly suitable for time series expression profiles. In view of this, we propose a novel imputation algorithm, which is specially suited for the estimation of missing values in gene expression time series data. The algorithm utilizes Dynamic Time Warping (DTW) distance in order to measure the similarity between time expression profiles, and subsequently selects for each gene expression profile with missing values a dedicated set of candidate profiles for estimation. Three different DTW-based imputation (DTWimpute) algorithms have been considered: position-wise, neighborhood-wise, and two-pass imputation. These have initially been prototyped in Perl, and their accuracy has been evaluated on yeast expression time series data using several different parameter settings. The experiments have shown that the two-pass algorithm consistently outperforms, in particular for datasets with a higher level of missing entries, the neighborhood-wise and the position-wise algorithms. The performance of the two-pass DTWimpute algorithm has further been benchmarked against the weighted K-Nearest Neighbors algorithm, which is widely used in the biological community; the former algorithm has appeared superior to the latter one. Motivated by these findings, indicating clearly the added value of the DTW techniques for missing value estimation in time series data, we have built an optimized C++ implementation of the two-pass DTWimpute algorithm. The software also provides for a choice between three different initial rough imputation methods.

Significance analysis of microarray transcript levels in time series experiments

BackgroundMicroarray time series studies are essential to understand the dynamics of molecular events. In order to limit the analysis to those genes that change expression over time, a first necessary step is to select differentially expressed transcripts. A variety of methods have been proposed to this purpose; however, these methods are seldom applicable in practice since they require a large number of replicates, often available only for a limited number of samples. In this data-poor context, we evaluate the performance of three selection methods, using synthetic data, over a range of experimental conditions. Application to real data is also discussed.ResultsThree methods are considered, to assess differentially expressed genes in data-poor conditions. Method 1 uses a threshold on individual samples based on a model of the experimental error. Method 2 calculates the area of the region bounded by the time series expression profiles, and considers the gene differentially expressed if the area exceeds a threshold based on a model of the experimental error. These two methods are compared to Method 3, recently proposed in the literature, which exploits splines fit to compare time series profiles. Application of the three methods to synthetic data indicates that Method 2 outperforms the other two both in Precision and Recall when short time series are analyzed, while Method 3 outperforms the other two for long time series.ConclusionThese results help to address the choice of the algorithm to be used in data-poor time series expression study, depending on the length of the time series.

A hidden Markov model-based approach for identifying timing differences in gene expression under different experimental factors

Time series experiments of cDNA microarrays have been commonly used in various biological studies and conducted under a lot of experimental factors. A popular approach of time series microarray analysis is to compare one gene with another in their expression profiles, and clustering expression sequences is a typical example. On the other hand, a practically important issue in gene expression is to identify the general timing difference that is caused by experimental factors. This type of difference can be extracted by comparing a set of time series expression profiles under a factor with those under another factor, and so it would be difficult to tackle this issue by using only a current approach for time series microarray analysis. We have developed a systematic method to capture the timing difference in gene expression under different experimental factors, based on hidden Markov models. Our model outputs a real-valued vector at each state and has a unique state transition diagram. The parameters of our model are trained from a given set of pairwise (generally multiplewise) expression sequences. We evaluated our model using synthetic as well as real microarray datasets. The results of our experiment indicate that our method worked favourably to identify the timing ordering under different experimental factors, such as that gene expression under heat shock tended to start earlier than that under oxidative stress. Supplementary data are available at Bioinformatics online.