George K. Acquaah-Mensah

Manual curation is not sufficient for annotation of genomic databases

MOTIVATION Knowledge base construction has been an area of intense activity and great importance in the growth of computational biology. However, there is little or no history of work on the subject of evaluation of knowledge bases, either with respect to their contents or with respect to the processes by which they are constructed. This article proposes the application of a metric from software engineering known as the found/fixed graph to the problem of evaluating the processes by which genomic knowledge bases are built, as well as the completeness of their contents. RESULTS Well-understood patterns of change in the found/fixed graph are found to occur in two large publicly available knowledge bases. These patterns suggest that the current manual curation processes will take far too long to complete the annotations of even just the most important model organisms, and that at their current rate of production, they will never be sufficient for completing the annotation of all currently available proteomes.

Transcription factor NRF2 regulates miR-1 and miR-206 to drive tumorigenesis

The mechanisms by which deregulated nuclear factor erythroid-2-related factor 2 (NRF2) and kelch-like ECH-associated protein 1 (KEAP1) signaling promote cellular proliferation and tumorigenesis are poorly understood. Using an integrated genomics and ¹³C-based targeted tracer fate association (TTFA) study, we found that NRF2 regulates miR-1 and miR-206 to direct carbon flux toward the pentose phosphate pathway (PPP) and the tricarboxylic acid (TCA) cycle, reprogramming glucose metabolism. Sustained activation of NRF2 signaling in cancer cells attenuated miR-1 and miR-206 expression, leading to enhanced expression of PPP genes. Conversely, overexpression of miR-1 and miR-206 decreased the expression of metabolic genes and dramatically impaired NADPH production, ribose synthesis, and in vivo tumor growth in mice. Loss of NRF2 decreased the expression of the redox-sensitive histone deacetylase, HDAC4, resulting in increased expression of miR-1 and miR-206, and not only inhibiting PPP expression and activity but functioning as a regulatory feedback loop that repressed HDAC4 expression. In primary tumor samples, the expression of miR-1 and miR-206 was inversely correlated with PPP gene expression, and increased expression of NRF2-dependent genes was associated with poor prognosis. Our results demonstrate that microRNA-dependent (miRNA-dependent) regulation of the PPP via NRF2 and HDAC4 represents a novel link between miRNA regulation, glucose metabolism, and ROS homeostasis in cancer cells.

The compositional structure of Gene Ontology terms.

An analysis of the term names in the Gene Ontology reveals the prevalence of substring relations between terms: 65.3% of all GO terms contain another GO term as a proper substring. This substring relation often coincides with a derivational relationship between the terms. For example, the term regulation of cell proliferation (GO:0042127) is derived from the term cell proliferation (GO:0008283) by addition of the phrase regulation of. Further, we note that particular substrings which are not themselves GO terms (e.g. regulation of in the preceding example) recur frequently and in consistent subtrees of the ontology, and that these frequently occurring substrings often indicate interesting semantic relationships between the related terms. We describe the extent of these phenomena--substring relations between terms, and the recurrence of derivational phrases such as regulation of--and propose that these phenomena can be exploited in various ways to make the information in GO more computationally accessible, to construct a conceptually richer representation of the data encoded in the ontology, and to assist in the analysis of natural language texts.

Network inference algorithms elucidate Nrf2 regulation of mouse lung oxidative stress

A variety of cardiovascular, neurological, and neoplastic conditions have been associated with oxidative stress, i.e., conditions under which levels of reactive oxygen species (ROS) are elevated over significant periods. Nuclear factor erythroid 2-related factor (Nrf2) regulates the transcription of several gene products involved in the protective response to oxidative stress. The transcriptional regulatory and signaling relationships linking gene products involved in the response to oxidative stress are, currently, only partially resolved. Microarray data constitute RNA abundance measures representing gene expression patterns. In some cases, these patterns can identify the molecular interactions of gene products. They can be, in effect, proxies for protein–protein and protein–DNA interactions. Traditional techniques used for clustering coregulated genes on high-throughput gene arrays are rarely capable of distinguishing between direct transcriptional regulatory interactions and indirect ones. In this study, newly developed information-theoretic algorithms that employ the concept of mutual information were used: the Algorithm for the Reconstruction of Accurate Cellular Networks (ARACNE), and Context Likelihood of Relatedness (CLR). These algorithms captured dependencies in the gene expression profiles of the mouse lung, allowing the regulatory effect of Nrf2 in response to oxidative stress to be determined more precisely. In addition, a characterization of promoter sequences of Nrf2 regulatory targets was conducted using a Support Vector Machine classification algorithm to corroborate ARACNE and CLR predictions. Inferred networks were analyzed, compared, and integrated using the Collective Analysis of Biological Interaction Networks (CABIN) plug-in of Cytoscape. Using the two network inference algorithms and one machine learning algorithm, a number of both previously known and novel targets of Nrf2 transcriptional activation were identified. Genes predicted as novel Nrf2 targets include Atf1, Srxn1, Prnp, Sod2, Als2, Nfkbib, and Ppp1r15b. Furthermore, microarray and quantitative RT-PCR experiments following cigarette-smoke-induced oxidative stress in Nrf2+/+ and Nrf2−/− mouse lung affirmed many of the predictions made. Several new potential feed-forward regulatory loops involving Nrf2, Nqo1, Srxn1, Prdx1, Als2, Atf1, Sod1, and Park7 were predicted. This work shows the promise of network inference algorithms operating on high-throughput gene expression data in identifying transcriptional regulatory and other signaling relationships implicated in mammalian disease.

Contrast and variability in gene names

We studied contrast and variability in a corpus of gene names to identify potential heuristics for use in performing entity identification in the molecular biology domain. Based on our findings, we developed heuristics for mapping weakly matching gene names to their official gene names. We then tested these heuristics against a large body of Medline abstracts, and found that using these heuristics can increase recall, with varying levels of precision. Our findings also underscored the importance of good information retrieval and of the ability to disambiguate between genes, proteins, RNA, and a variety of other referents for performing entity identification with high precision.

Pharmacokinetics of artemisinin delivered by oral consumption of Artemisia annua dried leaves in healthy vs. Plasmodium chabaudi-infected mice.

ETHNOPHARMACOLOGICAL RELEVANCE The Chinese have used Artemisia annua as a tea infusion to treat fever for >2000 years. The active component is artemisinin. Previously we showed that when compared to mice fed an equal amount of pure artemisinin, a single oral dose of dried leaves of Artemisia annua (pACT) delivered to Plasmodium chabaudi-infected mice reduced parasitemia at least fivefold. Dried leaves also delivered >40 times more artemisinin in the blood with no toxicity. The pharmacokinetics (PK) of artemisinin delivered from dried plant material has not been adequately studied. MATERIALS AND METHODS Healthy and Plasmodium chabaudi-infected mice were oral gavaged with pACT to deliver a 100 mg kg(-1) body weight dose of artemisinin. Concentrations of serum artemisinin and one of its liver metabolites, deoxyartemisinin, were measured over two hours by GCMS. RESULTS The first order elimination rate constant for artemisinin in pACT-treated healthy mice was estimated to be 0.80 h(-1) with an elimination half-life (T½) of 51.6 min. The first order absorption rate constant was estimated at 1.39 h(-1). Cmax and Tmax were 4.33 mg L(-1) and 60 min, respectively. The area under the curve (AUC) was 299.5 mg min L(-1). In contrast, the AUC for pACT-treated infected mice was significantly greater at 435.6 mg min L(-1). Metabolism of artemisinin to deoxyartemisinin was suppressed in infected mice over the period of observation. Serum levels of artemisinin in the infected mice continued to rise over the 120 min of the study period, and as a result, the T½ was not determined; the Cmax and Tmax were estimated at ≥6.64 mgL(-1) and ≥120 min, respectively. Groups of healthy mice were also fed either artemisinin or artemisinin mixed in mouse chow. When compared at 60 min, artemisinin was undetectable in the serum of mice fed 100 mg AN kg(-1) body weight. When plant material was present either as mouse chow or Artemisia annua pACT, artemisinin levels in the serum rose to 2.44 and 4.32 mg L(-1), respectively, indicating that the presence of the plant matrix, even that of mouse chow, had a positive impact on the appearance of artemisinin in the blood. CONCLUSIONS These results showed that artemisinin and one of its drug metabolites were processed differently in healthy and infected mice. The results have implications for possible therapeutic use of pACT in treating malaria and other artemisinin-susceptible diseases.

Ethanol sensitivity: a central role for CREB transcription regulation in the cerebellum

BackgroundLowered sensitivity to the effects of ethanol increases the risk of developing alcoholism. Inbred mouse strains have been useful for the study of the genetic basis of various drug addiction-related phenotypes. Inbred Long-Sleep (ILS) and Inbred Short-Sleep (ISS) mice differentially express a number of genes thought to be implicated in sensitivity to the effects of ethanol. Concomitantly, there is evidence for a mediating role of cAMP/PKA/CREB signalling in aspects of alcoholism modelled in animals. In this report, the extent to which CREB signalling impacts the differential expression of genes in ILS and ISS mouse cerebella is examined.ResultsA training dataset for Machine Learning (ML) and Exploratory Data Analyses (EDA) was generated from promoter region sequences of a set of genes known to be targets of CREB transcription regulation and a set of genes whose transcription regulations are potentially CREB-independent. For each promoter sequence, a vector of size 132, with elements characterizing nucleotide composition features was generated. Genes whose expressions have been previously determined to be increased in ILS or ISS cerebella were identified, and their CREB regulation status predicted using the ML scheme C4.5. The C4.5 learning scheme was used because, of four ML schemes evaluated, it had the lowest predicted error rate. On an independent evaluation set of 21 genes of known CREB regulation status, C4.5 correctly classified 81% of instances with F-measures of 0.87 and 0.67 respectively for the CREB-regulated and CREB-independent classes. Additionally, six out of eight genes previously determined by two independent microarray platforms to be up-regulated in the ILS or ISS cerebellum were predicted by C4.5 to be transcriptionally regulated by CREB. Furthermore, 64% and 52% of a cross-section of other up-regulated cerebellar genes in ILS and ISS mice, respectively, were deemed to be CREB-regulated.ConclusionThese observations collectively suggest that ethanol sensitivity, as it relates to the cerebellum, may be associated with CREB transcription activity.

Suppressed expression of T-box transcription factors is involved in senescence in chronic obstructive pulmonary disease.

Chronic obstructive pulmonary disease (COPD) is a major global health problem. The etiology of COPD has been associated with apoptosis, oxidative stress, and inflammation. However, understanding of the molecular interactions that modulate COPD pathogenesis remains only partly resolved. We conducted an exploratory study on COPD etiology to identify the key molecular participants. We used information-theoretic algorithms including Context Likelihood of Relatedness (CLR), Algorithm for the Reconstruction of Accurate Cellular Networks (ARACNE), and Inferelator. We captured direct functional associations among genes, given a compendium of gene expression profiles of human lung epithelial cells. A set of genes differentially expressed in COPD, as reported in a previous study were superposed with the resulting transcriptional regulatory networks. After factoring in the properties of the networks, an established COPD susceptibility locus and domain-domain interactions involving protein products of genes in the generated networks, several molecular candidates were predicted to be involved in the etiology of COPD. These include COL4A3, CFLAR, GULP1, PDCD1, CASP10, PAX3, BOK, HSPD1, PITX2, and PML. Furthermore, T-box (TBX) genes and cyclin-dependent kinase inhibitor 2A (CDKN2A), which are in a direct transcriptional regulatory relationship, emerged as preeminent participants in the etiology of COPD by means of senescence. Contrary to observations in neoplasms, our study reveals that the expression of genes and proteins in the lung samples from patients with COPD indicate an increased tendency towards cellular senescence. The expression of the anti-senescence mediators TBX transcription factors, chromatin modifiers histone deacetylases, and sirtuins was suppressed; while the expression of TBX-regulated cellular senescence markers such as CDKN2A, CDKN1A, and CAV1 was elevated in the peripheral lung tissue samples from patients with COPD. The critical balance between senescence and anti-senescence factors is disrupted towards senescence in COPD lungs.

Predicting the Subcellular Localization of Human Proteins Using Machine Learning and Exploratory Data Analysis

Identifying the subcellular localization of proteins is particularly helpful in the functional annotation of gene products. In this study, we use Machine Learning and Exploratory Data Analysis (EDA) techniques to examine and characterize amino acid sequences of human proteins localized in nine cellular compartments. A dataset of 3,749 protein sequences representing human proteins was extracted from the SWISS-PROT database. Feature vectors were created to capture specific amino acid sequence characteristics. Relative to a Support Vector Machine, a Multi-layer Perceptron, and a Naïve Bayes classifier, the C4.5 Decision Tree algorithm was the most consistent performer across all nine compartments in reliably predicting the subcellular localization of proteins based on their amino acid sequences (average Precision=0.88; average Sensitivity=0.86). Furthermore, EDA graphics characterized essential features of proteins in each compartment. As examples, proteins localized to the plasma membrane had higher proportions of hydrophobic amino acids; cytoplasmic proteins had higher proportions of neutral amino acids; and mitochondrial proteins had higher proportions of neutral amino acids and lower proportions of polar amino acids. These data showed that the C4.5 classifier and EDA tools can be effective for characterizing and predicting the subcellular localization of human proteins based on their amino acid sequences.

SAGA: A hybrid search algorithm for Bayesian Network structure learning of transcriptional regulatory networks

Bayesian Networks have been used for the inference of transcriptional regulatory relationships among genes, and are valuable for obtaining biological insights. However, finding optimal Bayesian Network (BN) is NP-hard. Thus, heuristic approaches have sought to effectively solve this problem. In this work, we develop a hybrid search method combining Simulated Annealing with a Greedy Algorithm (SAGA). SAGA explores most of the search space by undergoing a two-phase search: first with a Simulated Annealing search and then with a Greedy search. Three sets of background-corrected and normalized microarray datasets were used to test the algorithm. BN structure learning was also conducted using the datasets, and other established search methods as implemented in BANJO (Bayesian Network Inference with Java Objects). The Bayesian Dirichlet Equivalence (BDe) metric was used to score the networks produced with SAGA. SAGA predicted transcriptional regulatory relationships among genes in networks that evaluated to higher BDe scores with high sensitivities and specificities. Thus, the proposed method competes well with existing search algorithms for Bayesian Network structure learning of transcriptional regulatory networks.