MOTIVATION: Metabolomics plays an essential role in the growing systems biology approaches to unravel the relationships between metabolites and diseases. Liquid chromatography-mass spectrometry (LC-MS) is central to this effort because it can profile many metabolites from limited material. Yet, in a typical untargeted LC-MS-based metabolomics study, the majority of detected peaks remain unannotated, largely due to incomplete spectral libraries and uncertainties in peak picking, alignment, and the handling of isotopes and adducts. These limitations hinder seamless integration with other omics layers. RESULTS: We developed an AI-powered platform (aiSysMet) that uses statistical, machine learning, and deep learning methods for metabolomics data processing, metabolite annotation, and integrative analysis of multi-omics data. The platform's interactive and modular web interface allows users to easily build data analysis pipelines that can be executed in the cloud. AVAILABILITY: aiSysMet is freely available for non-commercial users on https://tools.omicscraft.com/aiSysMet.

Oncogenesis, a complex and multifaceted process, is profoundly modulated by miRNA's regulatory role in gene expression. Over the years, a substantial body of knowledge concerning miRNA and mRNA has been accumulated, drawing from both rigorous biological experiments and intricate statistical analyses. In the realm of statistical modeling, the integration of such information as "prior knowledge" often amplifies the model's ability to pinpoint molecular targets of significance. This study seeks to leverage prior knowledge of miRNA-mRNA regulatory interactions to map the dynamic landscape of interactions in the specific context of hepatocellular carcinoma (HCC). To address this, we introduce an evolved iteration of a Bayesian two-step integrative method previously established in the literature. This augmented approach includes improved computing efficiency when dealing with high dimensional data and a novel mechanistic submodel, which operates autonomously, devoid of prior knowledge. Employing this method, we identified two discrete gene lists: one informed by prior knowledge and the other independently inferred. This bifurcated strategy provides a comprehensive perspective on gene interactions. Our methodological advancement allows for a nuanced analysis of gene networks, distinguishing between direct and indirect gene relationships and considering miRNA influences with two available sub-mechanistic submodels. We introduce an approach to validate our findings using a biological interaction network, emphasizing the quality and relevance of identified gene-gene relationships. Metrics like the Matthews Correlation Coefficient (MCC) and the true discovery rate (TDR) further attest to the robustness of our findings. In summation, aside from improving the existing sub-mechanistic model that requires prior knowledge, this paper presents an innovative prior knowledge-free sub-mechanistic model as an alternative. It champions the use of biological networks for validation, underscoring the significance of methodological advancements in genomics research.

Hepatocellular carcinoma (HCC) has been an approved indication for the administration of immunotherapy since 2017, but biomarkers that predict therapeutic response have remained limited. Understanding and characterizing the tumor immune microenvironment enables better classification of these tumors and may reveal biomarkers that predict immunotherapeutic efficacy. In this paper, we applied a cell-type deconvolution algorithm using DNA methylation array data to investigate the composition of the tumor microenvironment in HCC. Using two publicly available datasets with a total cohort size of 57 patients, each with tumor and matched normal tissue samples, we identified key differences in immune cell composition. We found that NK cell abundance was significantly decreased in HCC tumors compared to adjacent normal tissue. We also applied DNA methylation "clocks" which estimate phenotypic aging and compared these findings to expression-based determinations of cellular senescence. Senescence and epigenetic aging was significantly increased in HCC tumors, and the degree of age acceleration and senescence was strongly associated with decreased NK cell abundance. In summary, we found that NK cell infiltration in the tumor microenvironment is significantly diminished, and that this loss of NK abundance is strongly associated with increased senescence and age-related phenotype. These findings point to key interactions between NK cells and the senescent tumor microenvironment and offer insights into the pathogenesis of HCC as well as potential biomarkers of therapeutic efficacy.

Compared to European Americans (EAs), the incidence of Hepatocellular Carcinoma (HCC) is higher in African Americans (AAs) and is associated with more advanced tumor stage at diagnosis and lower survival rates. Despite the importance of DNA methylation in neoplastic transformation, the mechanisms of epigenetic disparity between AAs and EAs are yet to be fully realized in HCC. In this paper we present an integrated analysis of genome-wide DNA methylation and gene expression in a racially diverse cohort of 16 HCC patients using the Illumina MethylationEPIC BeadChip and HiSeq 4000 platforms. A mixed-effects ANOVA model was applied to compare molecular changes between tumors and adjacent non-cancerous liver tissues. In an effort to control for underlying variation in tumor composition, we used DNA methylation to assess measurements of tumor purity and adjusted our analysis to account for potential bias due to cell mixture effects. As a result, we were able to identify 17,249 differentially methylated regions (DMRs) and 981 differentially methylated and expressed genes (DM+DE) associated with HCC. Further, 14,912 DMRs and 637 DM+DE genes were identified as being associated with the EA race and 3,257 DMRs and 156 DM+DE genes identified as being associated with the AA race in HCC. We were able to identify 20 candidate genes as potential tumor suppressors in HCC, including TBX15, IGF1R, WDR66, ITPKB, CFTR, KCNA3, CXCL12, EA-specific MT1L, PIK23R5, and AA-specific GHR and ADRA2B. In summary, we have identified differentially methylated and expressed markers of HCC that are specific to AA and EA racial groups. These findings offer insights into the molecular mechanisms of epigenetic regulation in HCC progression as well as identify potential markers to address racial disparity in precision therapy and diagnosis.

Metabolite identification is a major bottleneck in mass spectrometry-based metabolomic studies. In a typical untargeted metabolomic study that uses tandem MS (MS/MS), the identities of metabolites are determined by matching the experimental MS/MS data against those in spectral libraries. However, existing spectral libraries cover only a small fraction of known compounds. We introduce a neural network model to predict the identity of analytes whose reference measurements are not available in spectral libraries. Specifically, an artificial neural network model is trained to predict the fingerprints of compounds based on their experimental MS/MS spectra. Candidates are retrieved from metabolite databases based on molecular formula or precursor mass. The candidate with the most similar fingerprint is chosen as the predicted one. The method was evaluated via an independent dataset. The results show that our neural network model improves MS/MS-based metabolite identification accuracy by a considerable margin compared to previous computational methods, such as Metfrag and CSI:FingerID.

BACKGROUND: Conventional differential gene expression analysis by methods such as student's t-test, SAM, and Empirical Bayes often searches for statistically significant genes without considering the interactions among them. Network-based approaches provide a natural way to study these interactions and to investigate the rewiring interactions in disease versus control groups. In this paper, we apply weighted graphical LASSO (wgLASSO) algorithm to integrate a data-driven network model with prior biological knowledge (i.e., protein-protein interactions) for biological network inference. We propose a novel differentially weighted graphical LASSO (dwgLASSO) algorithm that builds group-specific networks and perform network-based differential gene expression analysis to select biomarker candidates by considering their topological differences between the groups. RESULTS: Through simulation, we showed that wgLASSO can achieve better performance in building biologically relevant networks than purely data-driven models (e.g., neighbor selection, graphical LASSO), even when only a moderate level of information is available as prior biological knowledge. We evaluated the performance of dwgLASSO for survival time prediction using two microarray breast cancer datasets previously reported by Bild et al. and van de Vijver et al. Compared with the top 10 significant genes selected by conventional differential gene expression analysis method, the top 10 significant genes selected by dwgLASSO in the dataset from Bild et al. led to a significantly improved survival time prediction in the independent dataset from van de Vijver et al. Among the 10 genes selected by dwgLASSO, UBE2S, SALL2, XBP1 and KIAA0922 have been confirmed by literature survey to be highly relevant in breast cancer biomarker discovery study. Additionally, we tested dwgLASSO on TCGA RNA-seq data acquired from patients with hepatocellular carcinoma (HCC) on tumors samples and their corresponding non-tumorous liver tissues. Improved sensitivity, specificity and area under curve (AUC) were observed when comparing dwgLASSO with conventional differential gene expression analysis method. CONCLUSIONS: The proposed network-based differential gene expression analysis algorithm dwgLASSO can achieve better performance than conventional differential gene expression analysis methods by integrating information at both gene expression and network topology levels. The incorporation of prior biological knowledge can lead to the identification of biologically meaningful genes in cancer biomarker studies.

In this paper, we introduce a computational method for constructing networks based on reverse phase protein array (RPPA) data to identify complex patterns in protein signaling. The method is applied to phosphoproteomic profiles of basal expression and activation/phosphorylation of 76 key signaling proteins in three breast cancer cell lines (MCF7, LCC1, and LCC9). Temporal RPPA data are acquired at 48h, 96h, and 144h after knocking down four genes in separate experiments. These genes are selected from a previous study as important determinants for breast cancer survival. Interaction networks between the proteins and phosphorylated proteins are constructed by analyzing the expression levels of protein pairs using a multivariate analysis of variance model. A new scoring criterion is introduced to determine the proteins that changed significantly at each of the three time points. Signaling networks are constructed based on statistically significant protein pairs selected from the RPPA data. Through a topology based analysis, we search for wiring patterns to help identify key network nodes that are associated with protein expression changes in various experimental conditions.

Liquid (or gas) chromatography coupled with mass spectrometry (LC-MS or GC-MS) allows quantitative comparison of biomolecular abundance in biological samples to help with the discovery of candidate biomarkers for complex diseases such as cancer. A fundamental challenge in using quantitative analysis of biomolecules by LC-MS or GC-MS for cancer biomarker discovery is owing to the heterogeneous nature of human biospecimens. Various contaminations present in cancerous tissues or adjacent non-cancerous constituents confound the characterization of molecular expression profiles and thus hinder the discovery of reliable biomarkers. We previously applied probabilistic purification model on a relatively small sample-size metabolomic data. In this study, we further apply probabilistic purification models on larger sample-size and multi-group metabolomic datasets acquired by analysis of liver tissues using both LC-MS and GC-MS. We demonstrate the advantages of incorporating purification models in retrieving underlying sources and reduce noise in metabolomic data. Furthermore, we investigate the benefit of the proposed models in improving our ability to detect changes in the level of metabolites among liver tissue from multiple groups (tumor, liver cirrhosis, and normal).

Studies associating changes in the levels of multiple biomolecules including proteins, glycans, glycoproteins, and metabolites with the onset of cancer have been widely investigated to identify clinically relevant diagnostic biomarkers. Advances in liquid or gas chromatography mass spectrometry (LC-MS, GC-MS) have enabled high-throughput qualitative and quantitative analysis of these biomolecules. While results from separate analyses of different biomolecules have been reported widely, the mutual information obtained by partly or fully combining them has been relatively unexplored. In this study, we investigate integrative analysis of proteins, N-glycans, and metabolites to take advantage of complementary information to improve the ability to distinguish cancer cases from controls. Specifically, support vector machine-recursive feature elimination algorithm is utilized to select a panel of proteins, N-glycans, and metabolites based on LC-MS and GC-MS data previously acquired by the analysis of blood samples from two cohorts in a liver cancer study. Improved performances are observed by integrative analysis compared to separate proteomic, glycomic, and metabolomic studies in distinguishing liver cancer cases from patients with liver cirrhosis.

Liquid (or gas) chromatography coupled with mass spectrometry (LC/GC-MS) allows quantitative comparison of biomolecular abundance in clinical samples to help with the discovery of candidate biomarkers for complex diseases such as cancer. A fundamental challenge in quantitation of biomolecules for cancer biomarker discovery is owing to the heterogeneous nature of clinical samples. Various contaminations from related disease tissues or adjacent non-cancerous constituents in a sample confound the characterization of molecular expression profiles and thus hinder the discovery of reliable biomarkers. This issue has been raised and discussed in analysis of microarray and RNA-seq data in cancer genomics studies. To the best of our knowledge, the issue has not yet been rigorously addressed in analyzing LC/GC-MS data that are generated in a variety of omic studies including proteomics and metabolomics. Purification of LC/GC-MS based biomolecular expression profiles is highly desired prior to subsequent analysis, e.g., quantitative comparison of the abundance of biomolecules in clinical samples. In this study, we applied a topic model to computationally deconvolute each of LC/GC-MS based cancer expression profiles and infer the underlying sample-specific pure cancer profiles. We demonstrated the capability of the model in capturing mixture proportions of contaminants and cancer profiles on a synthetic LC-MS dataset. Improved performances were also achieved on experimental LC-MS based serum proteomic and GC-MS based tissue metabolomic datasets acquired from patients with hepatocellular carcinoma (HCC).

Systems biology aims at unravelling the mechanisms of complex diseases by investigating how individual elements of the cell (e.g., genes, proteins, metabolites, etc.) interact with each other. Network-based methods provide an intuitive framework to model, characterize, and understand these interactions. To reconstruct a biological network, one can either query public databases for known interactions (knowledge-driven approach) or build a mathematical model to measure the associations from data (data-driven approach). In this paper, we propose a new network inference method, integrating knowledge and data-driven approaches. The method integrates prior biological knowledge (i.e., protein-protein interactions from BioGRID database) and a Gaussian graphical model (i.e., graphical LASSO algorithm) to construct robust and biologically relevant network. The network is then utilized to extract differential sub-networks between case and control groups using the result from a statistical analysis (e.g., logistic regression). We applied the proposed method on a proteomic dataset acquired by analysis of sera from hepatocellular carcinoma (HCC) cases and patients with liver cirrhosis. The differential sub-networks led to the identification of hub proteins and key pathways, whose relevance to HCC study has been confirmed by literature survey.

BACKGROUND: Gas chromatography coupled with mass spectrometry (GC-MS) is one of the technologies widely used for qualitative and quantitative analysis of small molecules. In particular, GC coupled to single quadrupole MS can be utilized for targeted analysis by selected ion monitoring (SIM). However, to our knowledge, there are no software tools specifically designed for analysis of GC-SIM-MS data. In this paper, we introduce a new R/Bioconductor package called SIMAT for quantitative analysis of the levels of targeted analytes. SIMAT provides guidance in choosing fragments for a list of targets. This is accomplished through an optimization algorithm that has the capability to select the most appropriate fragments from overlapping chromatographic peaks based on a pre-specified library of background analytes. The tool also allows visualization of the total ion chromatograms (TIC) of runs and extracted ion chromatograms (EIC) of analytes of interest. Moreover, retention index (RI) calibration can be performed and raw GC-SIM-MS data can be imported in netCDF or NIST mass spectral library (MSL) formats. RESULTS: We evaluated the performance of SIMAT using two GC-SIM-MS datasets obtained by targeted analysis of: (1) plasma samples from 86 patients in a targeted metabolomic experiment; and (2) mixtures of internal standards spiked in plasma samples at varying concentrations in a method development study. Our results demonstrate that SIMAT offers alternative solutions to AMDIS and MetaboliteDetector to achieve accurate detection of targets and estimation of their relative intensities by analysis of GC-SIM-MS data. CONCLUSIONS: We introduce a new R package called SIMAT that allows the selection of the optimal set of fragments and retention time windows for target analytes in GC-SIM-MS based analysis. Also, various functions and algorithms are implemented in the tool to: (1) read and import raw data and spectral libraries; (2) perform GC-SIM-MS data preprocessing; and (3) plot and visualize EICs and TICs.

We introduce a new method for normalization of data acquired by liquid chromatography coupled with mass spectrometry (LC-MS) in label-free differential expression analysis. Normalization of LC-MS data is desired prior to subsequent statistical analysis to adjust variabilities in ion intensities that are not caused by biological differences but experimental bias. There are different sources of bias including variabilities during sample collection and sample storage, poor experimental design, noise, etc. In addition, instrument variability in experiments involving a large number of LC-MS runs leads to a significant drift in intensity measurements. Although various methods have been proposed for normalization of LC-MS data, there is no universally applicable approach. In this paper, we propose a Bayesian normalization model (BNM) that utilizes scan-level information from LC-MS data. Specifically, the proposed method uses peak shapes to model the scan-level data acquired from extracted ion chromatograms (EIC) with parameters considered as a linear mixed effects model. We extended the model into BNM with drift (BNMD) to compensate for the variability in intensity measurements due to long LC-MS runs. We evaluated the performance of our method using synthetic and experimental data. In comparison with several existing methods, the proposed BNM and BNMD yielded significant improvement.

Recent technological advances have enabled the generation of various omic data (e.g., genomics, proteomics, metabolomics and glycomics) in a high-throughput manner. The integration of multi-omic data sets is desirable to unravel the complexity of a biological system. In this paper, we propose a new approach to investigate both inter and intra relationships for multi-omic data sets by using regularized canonical correlation analysis and graphical lasso. The application of this novel approach on real multi-omic data sets helps identify hub proteins and their neighbors that may be missed by typical statistical analysis to serve as biomarker candidates. Also, the integration of data from various cellular components (i.e., proteins, metabolites and glycans) offers the potential to discover more reliable biomarker candidates for complex disease.

Glycomics helps investigate the role glycosylation plays in complex diseases. Liquid chromatography (LC) coupled with mass spectrometry (MS) is routinely used to profile the glycans released from proteins in a biological sample. This enables us to compare observed glycans and their abundances among different biological samples to discover candidate biomarkers. One of the challenges in label-free LC/MS-based glycan profiling is the presence of various charge states and derived adduct ions. We propose a novel Glycan Profile Annotation (GPA) algorithm to automatically cluster and annotate these ions using a graphical model. Specifically, GPA aims to generate a list of unique neutral masses representing putative glycan composition derived from various charge states and multiple adducts. We demonstrate the performance of GPA in recognizing ions derived from the same glycan through analysis of LC/MS data from a serum biomarker discovery study. In addition, a simulation study is carried out to evaluate GPA's performance against existing tools in handling ambiguous cases.

One major challenge of systems biology is to infer biological networks. Classical graphical modeling methods that measure full conditional relationships between random variables may lead to unreliable results. This is partly due to the singular matrix problem when the number of variables exceeds the number of the samples. In this paper, we propose a low order partial correlation method to address this problem, trading off the small bias introduced by the low order constraint for the more reliable approximation of the network structure. Simulation results show that our proposed method works well under various conditions commonly seen in real applications and the inferred network faithfully uncovers the conditional independence relations among variables.

MOTIVATION: Liquid chromatography-mass spectrometry (LC-MS) has been widely used for profiling expression levels of biomolecules in various '-omic' studies including proteomics, metabolomics and glycomics. Appropriate LC-MS data preprocessing steps are needed to detect true differences between biological groups. Retention time (RT) alignment, which is required to ensure that ion intensity measurements among multiple LC-MS runs are comparable, is one of the most important yet challenging preprocessing steps. Current alignment approaches estimate RT variability using either single chromatograms or detected peaks, but do not simultaneously take into account the complementary information embedded in the entire LC-MS data. RESULTS: We propose a Bayesian alignment model for LC-MS data analysis. The alignment model provides estimates of the RT variability along with uncertainty measures. The model enables integration of multiple sources of information including internal standards and clustered chromatograms in a mathematically rigorous framework. We apply the model to LC-MS metabolomic, proteomic and glycomic data. The performance of the model is evaluated based on ground-truth data, by measuring correlation of variation, RT difference across runs and peak-matching performance. We demonstrate that Bayesian alignment model improves significantly the RT alignment performance through appropriate integration of relevant information. AVAILABILITY AND IMPLEMENTATION: MATLAB code, raw and preprocessed LC-MS data are available at http://omics.georgetown.edu/alignLCMS.html. CONTACT: [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

A Bayesian alignment model (BAM) is proposed for alignment of liquid chromatography-mass spectrometry (LC-MS) data. BAM belongs to the category of profile-based approaches, which are composed of two major components: a prototype function and a set of mapping functions. Appropriate estimation of these functions is crucial for good alignment results. BAM uses Markov chain Monte Carlo (MCMC) methods to draw inference on the model parameters and improves on existing MCMC-based alignment methods through 1) the implementation of an efficient MCMC sampler and 2) an adaptive selection of knots. A block Metropolis-Hastings algorithm that mitigates the problem of the MCMC sampler getting stuck at local modes of the posterior distribution is used for the update of the mapping function coefficients. In addition, a stochastic search variable selection (SSVS) methodology is used to determine the number and positions of knots. We applied BAM to a simulated data set, an LC-MS proteomic data set, and two LC-MS metabolomic data sets, and compared its performance with the Bayesian hierarchical curve registration (BHCR) model, the dynamic time-warping (DTW) model, and the continuous profile model (CPM). The advantage of applying appropriate profile-based retention time correction prior to performing a feature-based approach is also demonstrated through the metabolomic data sets.

Many complex diseases such as cancer are associated with changes in biological pathways and molecular networks rather than being caused by single gene alterations. A major challenge in the diagnosis and treatment of such diseases is to identify characteristic aberrancies in the biological pathways and molecular network activities and elucidate their relationship to the disease. This review presents recent progress in using high-throughput biological assays to decipher aberrant pathways and network activities. In particular, this review provides specific examples in which high-throughput data have been applied to identify relationships between diseases and aberrant pathways and network activities. The achievements in this field have been remarkable, but many challenges have yet to be addressed.

A Bayesian alignment model (BAM) is proposed for alignment of liquid chromatography-mass spectrometry (LC-MS) data. BAM is composed of two important components: prototype function and mapping function. Estimation of both functions is crucial for the alignment result. We use Markov chain Monte Carlo (MCMC) methods for inference of model parameters. To address the trapping effect in local modes, we propose a block Metropolis-Hastings algorithm that leads to better mixing behavior in updating the mapping function coefficients. We applied BAM to both simulated and real LC-MS datasets, and compared its performance with the Bayesian hierarchical curve registration model (BHCR). Performance evaluation on both simulated and real datasets shows satisfactory results in terms of correlation coefficients and ratio of overlapping peak areas.

Metabolite identification is the major bottle-neck in LC-MS based metabolomic investigations. The mass-based search approach often leaves a large fraction of metabolites with either no identification or multiple putative identifications. As manual verification of metabolites is laborious, computational approaches are needed to obtain more reliable putative identifications and prioritize them. In this paper, a computational pipeline is proposed to assist metabolite identification with improved coverage and prioritization capability. The pipeline is based on multiple pieces of publicly-available software and databases. The proposed pipeline is successfully applied in an LC-MS/MS based metabolomic study, where mass, retention time, and MS/MS spectrum were used to improve the accuracy of metabolite identification and to prioritize putative identifications for subsequent metabolite verification.

A novel framework of a probabilistic mixture regression model (PMRM) is presented for alignment of liquid chromatography-mass spectrometry (LC-MS) data with respect to retention time (RT) points. The expectation maximization algorithm is used to estimate the joint parameters of spline-based mixture regression models and prior transformation density models. The latter accounts for the variability in RT points and peak intensities. The applicability of PMRM for alignment of LC-MS data is demonstrated through three data sets. The performance of PMRM is compared with other alignment approaches including dynamic time warping, correlation optimized warping, and continuous profile model in terms of coefficient variation of replicate LC-MS runs and accuracy in detecting differentially abundant peptides/proteins.

Pancreatic cancer (PC) is the fourth leading cause of cancer death in the United States, with 4% survival 5 years after diagnosis. Patients with pancreatic cancer are usually diagnosed at late stages, when the disease is incurable. Sensitive and more specific biomarkers are thus critical for supporting new prevention, diagnostic or therapeutic strategies. Here, we report mass spectrometry-based metabolomic profiling of human pancreatic tumor and normal tissue. Multivariate data analysis shows significant alterations in the profiles of the tumor metabolome as compared to the normal. These findings offer an information-rich matrix for discovering novel biomarkers with potential for diagnostic or prognostic purposes.

In proteomics studies, liquid chromatography coupled to mass spectrometry (LC-MS) has proven to be a powerful technology to investigate differential expression of proteins/peptides that are characterized by their peak intensities, mass-to-charge ratio (m/z), and retention time (RT). The variable complexity of peptide mixtures and occasional drifts leads to substantial variations in m/z and RT dimensions. Thus, label-free differential protein expression studies by LC-MS technology require alignment with respect to both RT and m/z to ensure that same proteins/peptides are compared from multiple runs. In this study, we propose a new strategy to align LC-MALDI-TOF data by combining quality threshold cluster analysis and support vector regression. Our method performs alignment on the basis of measurements in three dimensions (RT, m/z, intensity). We demonstrate the suitability of our proposed method for alignment of LC-MALDI-TOF data through a previously published spike-in dataset and a new in-house generated spike-in dataset. A comparison of our method with other methods that utilize only RT and m/z dimensions reveals that the use of intensity measurements enhances alignment performance.

The availability of genome-wide biological network data opens up new possibilities to discover novel biomarkers and elucidate cancer-related complex mechanisms at network level. In this paper, we propose a novel module-based feature selection framework, which integrates biological network information and gene expression data to identify biomarkers, not as individual genes but as functional modules. Also, a large-scale analysis of ensemble feature selection concept is presented. The method allows combining features selected from multiple runs with various data subsampling to increase the reliability and classification accuracy of the final set of selected features. The results from four breast cancer studies demonstrate that the identified module biomarkers achieve: i) higher classification accuracy in independent validation datasets; ii) better reproducibility than individual gene biomarkers; iii) improved biological interpretability; and iv) enhanced enrichment in cancer-related “disease drivers”.

BACKGROUND: Integrating data from multiple global assays and curated databases is essential to understand the spatio-temporal interactions within cells. Different experiments measure cellular processes at various widths and depths, while databases contain biological information based on established facts or published data. Integrating these complementary datasets helps infer a mutually consistent transcriptional regulatory network (TRN) with strong similarity to the structure of the underlying genetic regulatory modules. Decomposing the TRN into a small set of recurring regulatory patterns, called network motifs (NM), facilitates the inference. Identifying NMs defined by specific transcription factors (TF) establishes the framework structure of a TRN and allows the inference of TF-target gene relationship. This paper introduces a computational framework for utilizing data from multiple sources to infer TF-target gene relationships on the basis of NMs. The data include time course gene expression profiles, genome-wide location analysis data, binding sequence data, and gene ontology (GO) information. RESULTS: The proposed computational framework was tested using gene expression data associated with cell cycle progression in yeast. Among 800 cell cycle related genes, 85 were identified as candidate TFs and classified into four previously defined NMs. The NMs for a subset of TFs are obtained from literature. Support vector machine (SVM) classifiers were used to estimate NMs for the remaining TFs. The potential downstream target genes for the TFs were clustered into 34 biologically significant groups. The relationships between TFs and potential target gene clusters were examined by training recurrent neural networks whose topologies mimic the NMs to which the TFs are classified. The identified relationships between TFs and gene clusters were evaluated using the following biological validation and statistical analyses: (1) Gene set enrichment analysis (GSEA) to evaluate the clustering results; (2) Leave-one-out cross-validation (LOOCV) to ensure that the SVM classifiers assign TFs to NM categories with high confidence; (3) Binding site enrichment analysis (BSEA) to determine enrichment of the gene clusters for the cognate binding sites of their predicted TFs; (4) Comparison with previously reported results in the literatures to confirm the inferred regulations. CONCLUSION: The major contribution of this study is the development of a computational framework to assist the inference of TRN by integrating heterogeneous data from multiple sources and by decomposing a TRN into NM-based modules. The inference capability of the proposed framework is verified statistically (e.g., LOOCV) and biologically (e.g., GSEA, BSEA, and literature validation). The proposed framework is useful for inferring small NM-based modules of TF-target gene relationships that can serve as a basis for generating new testable hypotheses.

MOTIVATION: Due to the large number of peaks in mass spectra of low-molecular-weight (LMW) enriched sera, a systematic method is needed to select a parsimonious set of peaks to facilitate biomarker identification. We present computational methods for matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) spectral data preprocessing and peak selection. In particular, we propose a novel method that combines ant colony optimization (ACO) with support vector machines (SVM) to select a small set of useful peaks. RESULTS: The proposed hybrid ACO-SVM algorithm selected a panel of eight peaks out of 228 candidate peaks from MALDI-TOF spectra of LMW enriched sera. An SVM classifier built with these peaks achieved 94% sensitivity and 100% specificity in distinguishing hepatocellular carcinoma from cirrhosis in a blind validation set of 69 samples. Area under the receiver operating characteristic (ROC) curve was 0.996. The classification capability of these peaks is compared with those selected by the SVM-recursive feature elimination method. AVAILABILITY: Supplementary material and MATLAB scripts to implement the methods described in this article are available at http://microarray.georgetown.edu/web/files/bioinf.htm. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

In this paper, we introduce a novel feature selection method that combines ant colony optimization (ACO) with support vector machine (SVM) to identify candidate biomarkers from mass spectral serum profiles. In addition, we present an innovative rule extraction algorithm that uses ACO to select accurate if-then rules for the classification of mass spectra. We applied the proposed feature selection and rule extraction methods to identify candidate biomarkers and extract if-then classification rules from MALDI-TOF spectra of enriched serum. The candidate biomarkers and the associated rules distinguished hepatocellular carcinoma patients from matched controls with high sensitivity and specificity

We present a novel algorithm that combines a recurrent neural network (RNN) and two swarm intelligence (SI) methods to infer a gene regulatory network (GRN) from time course gene expression data. The algorithm uses ant colony optimization (ACO) to identify the optimal architecture of an RNN, while the weights of the RNN are optimized using particle swarm optimization (PSO). Our goal is to construct an RNN whose response mimics gene expression data generated by time course DNA microarray experiments. We observed promising results in applying the proposed hybrid SI-RNN algorithm to infer networks of interaction from simulated and real-world gene expression data

A system incorporating a fuzzy c-means clustering and an ensemble of artificial neural networks (ANNs) is proposed to estimate chlorophyll-a (Chl a) concentration from remotely sensed reflectance (Rrs) measurements. The proposed method can be used to estimate Chl a concentration from Rrs measured at various locations representing heterogeneous water types. The performance of the proposed method is compared with the traditional approach, where a single ANN is used for all water types. We showed that the cluster-based approach has the potential to build a more global Chl a prediction model.

Serum profiling using mass spectrometry is an emerging technology with a great potential to provide biomarkers for complex diseases such as cancer. However, protein profiles obtained from current mass spectrometric technologies are characterized by their high dimensionality and complex spectra with substantial level of noise. These characteristics have generated challenges in discovery of proteins and protein-profiles that distinguish cancer patients from healthy individuals. This paper proposes a novel machine learning method that combines support vector machines with particle swarm optimization for biomarker discovery. Prior to applying the proposed biomarker selection algorithm, low-level analysis methods are used for smoothing, baseline correction, normalization, and peak detection. The proposed method is applied for biomarker discovery from serum mass spectral profiles of liver cancer patients and controls.

MOTIVATION: Mass spectrometric profiles of peptides and proteins obtained by current technologies are characterized by complex spectra, high dimensionality and substantial noise. These characteristics generate challenges in the discovery of proteins and protein-profiles that distinguish disease states, e.g. cancer patients from healthy individuals. We present low-level methods for the processing of mass spectral data and a machine learning method that combines support vector machines, with particle swarm optimization for biomarker selection. RESULTS: The proposed method identified mass points that achieved high prediction accuracy in distinguishing liver cancer patients from healthy individuals in SELDI-QqTOF profiles of serum. AVAILABILITY: MATLAB scripts to implement the methods described in this paper are available from the HWR's lab website http://lombardi.georgetown.edu/labpage

Inversion of ocean color reflectance measurements can be cast as an optimization problem, where certain parameters of a forward model are optimized in order to make the forward modelled spectral reflectance match the spectral reflectance of a given in situ sample. A simulated ocean color dataset is used to test the capability of a recently introduced global optimization process, particle swarm optimization (PSO), compared to the more common genetic algorithm technique. On the test data, the PSO method has been shown to outperform the genetic algorithms (GA) in terms of model error. Also important to the ocean color (often remotely sensed) problem is the speed advantage which PSO affords over GA.

A novel approach is presented for determining confidence values for bases called in DNA sequencing. The approach employs fuzzy logic, which forwards flexibility, adaptability and intuition through the use of linguistic variables and fuzzy membership. The proposed fuzzy logic-based method determines the confidence values based on the height, the peakness, and the spacing of the first most likely candidate (the base called) and the peakness and height for the second likely candidate. The technique has been tested on over 3000 ABI 3700 DNA files and the result has shown improved performance over the existing Phred's and ABTs quality value.

Light availability to seagrasses is a major criterion limiting the distribution of seagrasses. Decreased water clarity and resulting reduced light penetration have been cited as major factors responsible for the decline in seagrasses. Light attenuation coefficient is an important parameter that indicates the light attenuated by the water column and can thereby be an indicator of seagrass health. Though, in practice, linear light attenuation models have been commonly used, there is a need for a more accurate model that can take into account the non-linearities present in coastal and estuarine environments. This paper presents neural network-based light attenuation models for monitoring the seagrass health in the Indian River Lagoon, FL. For performance evaluation, results of the developed neural network models are compared with linear regression models, model trees, and support vector machines.

Inversion of ocean color reflectance measurements can be cast as an optimization problem, where particular parameters of a forward model are optimized in order to make the forward-modeled spectral reflectance match the spectral reflectance of a given in situ sample. Here, a simulated ocean color dataset is used to test the capability of a recently introduced global optimization process, particle swarm optimization (PSO), in the retrieval of optical properties from ocean color. The performance of the PSO method was compared with the more common genetic algorithms (GA) in terms of model accuracy and computation time. The PSO method has been shown to outperform the GA in terms of model error. Of particular importance to ocean color remote sensing is the speed advantage that PSO affords over GA.

DNA microarray technology enables a parallel analysis of the expression of genes in an organism. The wealth of spatio-temporal data generated by this technology allows researchers to potentially reverse engineer the genetic network. Fuzzy logic has been proposed as a method to analyze the relationships between genes. This method can identify interacting genes that fit a known fuzzy model of gene interaction by testing all combinations of gene expression profiles. However, this approach is slow and computationally complex. This paper introduces improvements made in terms of reducing computation time and generalizing the gene regulatory model to accommodate co-activators and co-repressors. Improvement in computation time is achieved by using clustering as a pre-processing method, thereby reducing the total number of gene combinations analyzed. This will allow the algorithm to run in a shorter amount of time with minimal effect on the results. The proposed technique will pave the way towards the creation of a generalized gene interaction model that can accommodate any combination of genes.

Monitoring seagrass health gives vital clues about the estuarine water quality, which is crucial for the existence of many aquatic plants and animals. Photosynthetic efficiency is a measure of plant stress and can be used to monitor seagrass health. However, insitu measurements of photosynthetic efficiency are time consuming and expensive. In this paper, neural network-based models are developed to estimate photosynthetic efficiency of a seagrass species, Zostera capricorni, from spectral reflectance measurements. The proposed neural network-based approach can be extended for other seagrass species by combining an ensemble of neural networks with a classifier. After identifying the type of seagrass species using the classifier, the neural network model that corresponds to the identified species is used to estimate photosynthetic efficiency.

This paper presents a novel clustering technique known as adaptive double self-organizing map (ADSOM). ADSOM has a flexible topology and perform clustering and cluster visualization simultaneously, thereby requiring no a priori knowledge about the number of clusters. ADSOM combines features of the popular self-organizing map (SOM) with two-dimensional position vectors, which serve as a visualization tool to accurately determine the number of clusters present in the data. ADSOM updates its free parameters during training and it allows convergence of its position vectors to a fairly consistent number of clusters provided that its initial number of nodes is greater than the expected number of clusters. A novel index is introduced based on hierarchical clustering of the final locations of position vectors. The index allows automatic detection of the number of clusters, thereby reducing human error that could be incurred from counting cluster visually. The reliance of ADSOM in identifying the number of clusters is proven by applying it to publicly available yeast gene expression data.

In this paper, neural network-based methods incorporating ensemble learning techniques are presented that estimate chlorophyll /spl alpha/ (chl /spl alpha/) concentration in the coastal waters of the Gulf of Maine (GOM). A dataset was constructed consisting of in situ chl measurements from the GOM matched with satellite data from the sea-viewing wide-field-of-view sensor (SeaWiFS). These data were used to develop models using diverse neural network ensembles for estimation of chl /spl alpha/ concentration from satellite-retrieved ocean reflectances. Results indicate that the models are able to generalize across geographical and temporal variation, and are resilient to uncertainty such as that introduced by poor atmospheric correction, or radiance contributions from non-chl /spl alpha/ components in case 2 waters.

Seagrasses are marine plants that provide many services such as primary productivity, food web interactions, shelter, nutrient cycling and habitat stabilization that are essential to marine and estuarine ecosystems. Therefore, monitoring seagrass health is crucial for the existence of many marine aquatic plants and animals. The minimal light requirement of seagrasses is about 20-30% of the total light measured just below the surface. This is relatively high compared to terrestrial plants and phytoplankton, which underlines the importance of water transparency for these species. Hence, light penetration into estuarine waters is critical for seagrass survival. In this paper, an approach to estimate the light attenuation coefficient from water quality parameters using neural networks is proposed. The model is compared with linear regression models such as step-wise linear regression and linear least squares regression. The light attenuation model presented here can be used for monitoring water quality and thereby seagrass health.

In an industrial nonwoven production process, data related to spun-web quality including firmness, thickness, and weight are measured online. However, these measurements provide insufficient information about the spun-web quality. As a result, other parameters that give a better insight into the quality of the product have been employed. One of these parameters reveals the minimum force necessary to pull apart a given size of piece of spun-web. It is measured in a quality control laboratory by stretching the ends of a sample piece of spun-web to a high degree of tension. Since the laboratory measurements are infrequent and off-line, an on-line monitoring of product quality can hardly be carried out. Neural network-based virtual sensors are developed based on historical data that incorporate both online and off-line measurements. The virtual sensors provide estimates of key quality parameters at the measurement frequency of the on-line measurements. They are implemented at a production facility of Freudenberg Nonwovens KG in Germany and have been successfully employed for on-line quality monitoring.

Compression plants are characterized by the fact that they possess a large number of state variables that are coupled non-linearly with each other. The paper focuses on the exploration of recurrent neural networks for modeling the dynamic behavior of a laboratory setup of a compression plant. Using data collected from the setup, recurrent multilayer perceptron networks are trained. The networks are validated not only with test data measured under similar external conditions but also with those that are gathered when the measurements of the external temperature are beyond the range inspected during the collection of the training data. Despite a significant change in external conditions, the validation results showed a fairly good performance in a multi-step prediction of the temperature and relative humidity inside the refrigerator.