Prospective Validation of ThyroSPEC Molecular Testing of Indeterminate Thyroid Nodule Cytology Following Diagnostic Pathway Optimization.

Background: Molecular testing for cytologically indeterminate thyroid nodules (ITNs) is often reported with incomplete data on clinical assessment and ultrasound malignancy risk (USMR) stratification. This study aimed to clinically validate the diagnostic accuracy of a novel molecular test, assess the incremental preoperative malignancy risk of other clinical factors, and measure the impacts of introducing molecular testing at the population level. Methods: Comprehensive clinical data were collected prospectively for the first 615 consecutive patients with ITNs in a centralized health care system following implementation of a reflexive molecular test. Clinical data include patient history, method of nodule discovery, clinical assessment, USMR, cytology, molecular testing, and surgery or follow-up along with surgeon notes on surgical decision-making. Accuracy of molecular testing and the impact of the introduction of molecular testing were calculated. A multivariable regression model was developed to identify which clinical factors have the most diagnostic significance for ITNs. Results: A locally developed, low-cost molecular test achieved a negative predictive value (NPV) of 76-91% [confidence interval, CI 66-95%] and a positive predictive value (PPV) of 46-65% [CI 37-75%] in ITNs using only residual material from standard liquid cytology fine-needle aspiration (FNA). Sensitivity was highest (80%; [CI 63-92%]) in the American Thyroid Association (ATA) intermediate-suspicion ultrasound category, and lowest (46%; [CI 19-75%]) in the ATA high-suspicion ultrasound category. Following implementation of molecular testing, diagnostic yield increased by 14% (p = 0.2442) and repeat FNAs decreased by 24% (p = 0.05). Mutation was the primary reason for surgery in 76% of resected, mutation-positive patients. High-risk mutations were associated with a 58% (p = 0.0001) shorter wait for surgery. Twenty-six percent of patients with a negative molecular test result underwent surgery. Multivariable regression highlighted molecular testing and USMR as significantly associated with malignancy. Conclusions: Molecular testing improves preoperative risk stratification but requires further stratification for intermediate-risk mutations. Incorporation of clinical factors (especially USMR) with molecular testing may increase the sensitivity for detection of malignancy. Introduction of molecular testing offers some clinical benefits even in a low resection rate setting, and directly influences surgical decision-making. This study illustrates the importance of the local diagnostic pathway in ensuring appropriate integrated use of molecular testing for best outcomes.

CACTUS: cancer image annotating, calibrating, testing, understanding and sharing in breast cancer histopathology.

OBJECTIVE: Develop CACTUS (cancer image annotating, calibrating, testing, understanding and sharing) as a novel web application for image archiving, annotation, grading, distribution, networking and evaluation. This helps pathologists to avoid unintended mistakes leading to quality assurance, teaching and evaluation in anatomical pathology. Effectiveness of the tool has been demonstrated by assessing pathologists performance in the grading of breast carcinoma and by comparing inter/intra-observer assessment of grading criteria amongst pathologists reviewing digital breast cancer images. Reproducibility has been assessed by inter-observer (kappa statistics) and intra-observer (intraclass correlation coefficient) concordance rates. RESULTS: CACTUS has been evaluated using a surgical pathology application-the assessment of breast cancer grade. We used CACTUS to present standardized images to four pathologists of differing experience. They were asked to evaluate all images to determine their assessment of Nottingham grade of a series of breast carcinoma cases. For each image, they were asked for their overall grade impression. CACTUS helps and guides pathologists to improve disease diagnosis with higher confidence and thereby reduces their workload and bias. CACTUS can be useful for both disseminating anatomical pathology images for teaching, as well as for evaluating agreement amongst pathologists or against a gold standard for evaluation or quality assurance.

Newly recognized non-adenomatous lesions associated with enteric carcinomas in inflammatory bowel disease - Report of six rare and unique cases.

It is the current view that the inflammatory bowel disease (IBD)-associated precancerous lesions may be adenomatous (with classic cytologic dysplasia) and non-adenomatous (without frank cytologic dysplasia), and the latter ones are in various histomorphologies including serrated, mucinous, eosinophilic (goblet cell deficient), and differentiated (dysplasia with terminal epithelial differentiation) types. By retrospectively reviewing the surgically resected IBD-associated colorectal and ileal carcinomas (×53), analyzing the background epithelial changes/lesions in the mucosa surrounding and adjacent to invasive carcinomas, and testing the key molecular profile (KRAS, BRAF, PIK3CA, NRAS, p53, mismatch repair proteins, and SAT-B2) known to be involved in colorectal carcinogenesis, we identified 6 representative, rare and unique cases, in which non-adenomatous lesions were clearly in vicinity and in transition to invasive carcinomas. Furthermore, we identified certain colonic carcinoma-related molecular alterations, and thus further confirmed the neoplastic nature of various non-adenomatous lesions. It was also revealed that non-adenomatous lesions are heterogeneous in both morphology and molecular alterations, and that it is common to have more than one type of lesions be associated with a carcinoma. Moreover, mixed focal adenomatous dysplasia was common, which may be the necessary step in the malignant transformation of the non-adenomatous lesions.

Histological and molecular diversity and heterogeneity of precancerous lesions associated with inflammatory bowel diseases.

AIMS: Inflammatory bowel disease (IBD)-associated precancerous lesions may be adenomatous or non-adenomatous with various histomorphologies. We aim to validate the newly proposed classification, to explore the neoplastic nature of the non-adenomatous lesions and to elucidate the molecular mechanisms underlying the different histomorphologies. METHODS: 44 background precursor lesions identified in 53 cases of surgically resected IBD-associated colorectal and ileal carcinomas were reviewed for the histomorphological features (classified into adenomatous, mucinous, sessile serrated adenoma (SSA)-like, traditional serrated adenoma-like, differentiated, eosinophilic and serrated not otherwise specified (NOS)) and analysed for a key panel of colonic cancer-related molecular markers. RESULTS: Approximately 60% of the lesions were adenomatous, of which some had mixed serrated, mucinous or eosinophilic changes. The remaining non-adenomatous lesions, including all other types except SSA-like type, mostly showed mixed features and focal adenomatous dysplasia. KRAS mutation and p53 mutant-type expression were found in about half cases across all types, while PIK3CA mutation only in some of adenomatous and eosinophilic lesions and MLH1/PMS2 loss in a subset of adenomatous, mucinous and eosinophilic but not in differentiated and serrated lesions. SAT-B2 or PTEN loss and IMP3 overexpression were seen in a small subset of lesions. No BRAF, NRAS or EGFR gene mutation was detected in any type. Certain molecular-morphological correlations were demonstrated; however, no single or combined molecular alteration(s) was specific to any particular morphological type. CONCLUSIONS: IBD-associated precancerous lesions are heterogeneous both histologically and molecularly. True colitis-associated adenomatous lesions are unlikely conventional adenomas. Non-adenomatous lesions without frank cytologic dysplasia should also be regarded as neoplastic.

Primary Pleural Squamous Cell Carcinoma in a Free-Ranging River Otter (Lontra canadensis).

An adult female North American river otter (Lontra canadensis) presented with multiple intrathoracic masses identified histologically as squamous cell carcinoma. Immunohistochemical staining patterns for high- molecular-weight keratin, p40, p63, calretinin, and TTF-1, along with the gross and histologic findings, indicated a primary pleural squamous cell carcinoma as the most likely diagnosis.

BreCaHAD: a dataset for breast cancer histopathological annotation and diagnosis.

OBJECTIVES: Histopathological tissue analysis by a pathologist determines the diagnosis and prognosis of most tumors, such as breast cancer. To estimate the aggressiveness of cancer, a pathologist evaluates the microscopic appearance of a biopsied tissue sample based on morphological features which have been correlated with patient outcome. DATA DESCRIPTION: This paper introduces a dataset of 162 breast cancer histopathology images, namely the breast cancer histopathological annotation and diagnosis dataset (BreCaHAD) which allows researchers to optimize and evaluate the usefulness of their proposed methods. The dataset includes various malignant cases. The task associated with this dataset is to automatically classify histological structures in these hematoxylin and eosin (H&E) stained images into six classes, namely mitosis, apoptosis, tumor nuclei, non-tumor nuclei, tubule, and non-tubule. By providing this dataset to the biomedical imaging community, we hope to encourage researchers in computer vision, machine learning and medical fields to contribute and develop methods/tools for automatic detection and diagnosis of cancerous regions in breast cancer histology images.

Molecular Auditing: An Evaluation of Unsuspected Tissue Specimen Misidentification.

CONTEXT.­: Specimen misidentification is the most significant error in laboratory medicine, potentially accounting for hundreds of millions of dollars in extra health care expenses and significant morbidity in patient populations in the United States alone. New technology allows the unequivocal documentation of specimen misidentification or contamination; however, the value of this technology currently depends on suspicion of the specimen integrity by a pathologist or other health care worker. OBJECTIVE.­: To test the hypothesis that there is a detectable incidence of unsuspected tissue specimen misidentification among cases submitted for routine surgical pathology examination. DESIGN.­: To test this hypothesis, we selected specimen pairs that were obtained at different times and/or different hospitals from the same patient, and compared their genotypes using standardized microsatellite markers used commonly for forensic human DNA comparison in order to identify unsuspected mismatches between the specimen pairs as a trial of "molecular auditing." We preferentially selected gastrointestinal, prostate, and skin biopsies because we estimated that these types of specimens had the greatest potential for misidentification. RESULTS.­: Of 972 specimen pairs, 1 showed an unexpected discordant genotype profile, indicating that 1 of the 2 specimens was misidentified. To date, we are unable to identify the etiology of the discordance. CONCLUSIONS.­: These results demonstrate that, indeed, there is a low level of unsuspected tissue specimen misidentification, even in an environment with careful adherence to stringent quality assurance practices. This study demonstrates that molecular auditing of random, routine biopsy specimens can identify occult misidentified specimens, and may function as a useful quality indicator.

Assessing the performance of a Loop Mediated Isothermal Amplification (LAMP) assay for the detection and subtyping of high-risk suptypes of Human Papilloma Virus (HPV) for Oropharyngeal Squamous Cell Carcinoma (OPSCC) without DNA purification.

BACKGROUND: Oropharyngeal Squamous Cell Carcinoma (OPSCC) is increasing in incidence despite a decline in traditional risk factors. Human Papilloma Virus (HPV), specifically subtypes 16, 18, 31 and 35, has been implicated as the high-risk etiologic agent. HPV positive cancers have a significantly better prognosis than HPV negative cancers of comparable stage, and may benefit from different treatment regimens. Currently, HPV related carcinogenesis is established indirectly through Immunohistochemistry (IHC) staining for p16, a tumour suppressor gene, or polymerase chain reaction (PCR) that directly tests for HPV DNA in biopsied tissue. Loop mediated isothermal amplification (LAMP) is more accurate than IHC, more rapid than PCR and is significantly less costly. In previous work we showed that a subtype specific HPV LAMP assay performed similar to PCR on purified DNA. In this study we examined the performance of this LAMP assay without DNA purification. METHODS: We used LAMP assays using established primers for HPV 16 and 18, and new primers for HPV 31 and 35. LAMP reaction conditions were tested on serial dilutions of plasmid HPV DNA to confirm minimum viral copy number detection thresholds. LAMP was then performed directly on different human cell line samples without DNA purification. RESULTS: Our LAMP assays could detect 105, 103, 104, and 105 copies of plasmid DNA for HPV 16, 18, 31, and 35, respectively. All primer sets were subtype specific, with no cross-amplification. Our LAMP assays also reliably amplified subtype specific HPV DNA from samples without requiring DNA isolation and purification. CONCLUSIONS: The high risk OPSCC HPV subtype specific LAMP primer sets demonstrated, excellent clinically relevant, minimum copy number detection thresholds with an easy readout system. Amplification directly from samples without purification illustrated the robust nature of the assay, and the primers used. This lends further support HPV type specific LAMP assays, and these specific primer sets and assays can be further developed to test for HPV in OPSCC in resource and lab limited settings, or even bedside testing.

WSB1: from homeostasis to hypoxia.

The wsb1 gene has been identified to be important in developmental biology and cancer. A complex transcriptional regulation of wsb1 yields at least three functional transcripts. The major expressed isoform, WSB1 protein, is a substrate recognition protein within an E3 ubiquitin ligase, with the capability to bind diverse targets and mediate ubiquitinylation and proteolytic degradation. Recent data suggests a new role for WSB1 as a component of a neuroprotective pathway which results in modification and aggregation of neurotoxic proteins such as LRRK2 in Parkinson's Disease, via an unusual mode of protein ubiquitinylation.WSB1 is also involved in thyroid hormone homeostasis, immune regulation and cellular metabolism, particularly glucose metabolism and hypoxia. In hypoxia, wsb1 is a HIF-1 target, and is a regulator of the degradation of diverse proteins associated with the cellular response to hypoxia, including HIPK2, RhoGDI2 and VHL. Major roles are to both protect HIF-1 function through degradation of VHL, and decrease apoptosis through degradation of HIPK2. These activities suggest a role for wsb1 in cancer cell proliferation and metastasis. As well, recent work has identified a role for WSB1 in glucose metabolism, and perhaps in mediating the Warburg effect in cancer cells by maintaining the function of HIF1. Furthermore, studies of cancer specimens have identified dysregulation of wsb1 associated with several types of cancer, suggesting a biologically relevant role in cancer development and/or progression.Recent development of an inducible expression system for wsb1 could aid in the further understanding of the varied functions of this protein in the cell, and roles as a potential oncogene and neuroprotective protein.

Integrating text mining, data mining, and network analysis for identifying genetic breast cancer trends.

BACKGROUND: Breast cancer is a serious disease which affects many women and may lead to death. It has received considerable attention from the research community. Thus, biomedical researchers aim to find genetic biomarkers indicative of the disease. Novel biomarkers can be elucidated from the existing literature. However, the vast amount of scientific publications on breast cancer make this a daunting task. This paper presents a framework which investigates existing literature data for informative discoveries. It integrates text mining and social network analysis in order to identify new potential biomarkers for breast cancer. RESULTS: We utilized PubMed for the testing. We investigated gene-gene interactions, as well as novel interactions such as gene-year, gene-country, and abstract-country to find out how the discoveries varied over time and how overlapping/diverse are the discoveries and the interest of various research groups in different countries. CONCLUSIONS: Interesting trends have been identified and discussed, e.g., different genes are highlighted in relationship to different countries though the various genes were found to share functionality. Some text analysis based results have been validated against results from other tools that predict gene-gene relations and gene functions.

Multi-scale modularity and motif distributional effect in metabolic networks.

Metabolism is a set of fundamental processes that play important roles in a plethora of biological and medical contexts. It is understood that the topological information of reconstructed metabolic networks, such as modular organization, has crucial implications on biological functions. Recent interpretations of modularity in network settings provide a view of multiple network partitions induced by different resolution parameters. Here we ask the question: How do multiple network partitions affect the organization of metabolic networks? Since network motifs are often interpreted as the super families of evolved units, we further investigate their impact under multiple network partitions and investigate how the distribution of network motifs influences the organization of metabolic networks. We studied Homo sapiens, Saccharomyces cerevisiae and Escherichia coli metabolic networks; we analyzed the relationship between different community structures and motif distribution patterns. Further, we quantified the degree to which motifs participate in the modular organization of metabolic networks.

Evaluating predictive performance of network biomarkers with network structures.

Network is a powerful structure which reveals valuable characteristics of the underlying data. However, previous work on evaluating the predictive performance of network-based biomarkers does not take nodal connectedness into account. We argue that it is necessary to maximize the benefit from the network structure by employing appropriate techniques. To address this, we aim to learn a weight coefficient for each node in the network from the quantitative measure such as gene expression data. The weight coefficients are computed from an optimization problem which minimizes the total weighted difference between nodes in a network structure; this can be expressed in terms of graph Laplacian. After obtaining the coefficient vector for the network markers, we can then compute the corresponding network predictor. We demonstrate the effectiveness of the proposed method by conducting experiments using published breast cancer biomarkers with three patient cohorts. Network markers are first grouped based on GO terms related to cancer hallmarks. We compare the predictive performance of each network marker group across gene expression datasets. We also evaluate the network predictor against the average method for feature aggregation. The reported results show that the predictive performance of network markers is generally not consistent across patient cohorts.

B-lymphoblastic leukemia/lymphoma: overexpression of nuclear DNA repair protein PARP-1 correlates with antiapoptotic protein Bcl-2 and complex chromosomal abnormalities.

Poly(ADP-ribose) polymerase-1 (PARP-1) and Bcl-2 are emerging as therapeutic targets in various cancers. The former is a DNA repair protein associated with genomic stability and apoptosis, whereas the latter is an antiapoptotic protein having a DNA repair function through inhibition of PARP-1. Because genomic stability is critical for prognosis in B-lymphoblastic leukemia/lymphoma (B-ALL), we studied the expression of PARP-1 and Bcl-2 proteins in patients with B-ALL of different ages and compared the results with cytogenetic data. The PARP-1 protein was overexpressed in about two-thirds (61%) of patients with B-ALL. It had a nuclear location, whereas Bcl-2 protein was cytosolic. Expression of the 2 proteins showed a highly positive correlation (ρ = 0.367; P < .001). Overexpression of PARP-1 correlated with a complex karyotype (P = .030), and this correlation remained significant for coexpression of PARP-1 and Bcl-2 proteins (χ(2) = 7.498; P = .024) as well as after exclusion of pediatric patients (n = 9, P = .042). Overexpression of PARP-1 was not significantly more common in diploid versus aneuploid karyotypes (50% versus 59%, P = .610). The PARP-1 protein showed no correlation with specific chromosomal abnormalities associated with prognosis in B-ALL, as defined by the World Health Organization. In conclusion, high expression of the PARP-1 protein among patients with B-ALL is related to a complex karyotype and Bcl-2 positivity. Although these findings require validation in a larger population, the observations will be valuable in planning therapeutic trials (such as of PARP inhibitors and BH3 mimetics).

Expression of MS4A and TMEM176 Genes in Human B Lymphocytes.

The MS4A gene family in humans includes CD20 and at least 15 other genes. CD20 exists as homo-oligomers in the plasma membrane, however different MS4A proteins expressed in the same cell may hetero-oligomerize. Given the importance of CD20 in B-cell function and as a therapeutic target, we sought to explore the potential for CD20 hetero-oligomerization with other MS4A proteins. We investigated expression in primary human B-cells of the four MS4A genes previously shown to be expressed in human B-cell lines (MS4A4A, MS4A6A, MS4A7, MS4A8B), as well as two genes comprising the closely related TMEM176 gene family, with a view to identifying candidates for future investigation at the protein level. TMEM176A and TMEM176B transcripts were either not detected, or were detected at relatively low levels in a minority of donor B-cell samples. MS4A4A and MS4A8B transcripts were not detected in any normal B-cell sample. MS4A6A and MS4A7 transcripts were detected at low levels in most samples, however the corresponding proteins were not at the plasma membrane when expressed as GFP conjugates in BJAB cells. We also examined expression of these genes in chronic lymphocytic leukemia (CLL), and found that it was similar to normal B-cells with two exceptions. First, whereas MS4A4A expression was undetected in normal B-cells, it was expressed in 1/14 CLL samples. Second, compared to expression levels in normal B-cells, MS4A6A transcripts were elevated in 4/14 CLL samples. In summary, none of the MS4A/TMEM176 genes tested was expressed at high levels in normal or in most CLL B-cells. MS4A6A and MS4A7 were expressed at low levels in most B-cell samples, however the corresponding proteins may not be positioned at the plasma membrane. Altogether, these data suggest that CD20 normally does not form hetero-oligomers with other MS4A proteins and that there are unlikely to be other MS4A proteins in CLL that might provide useful alternate therapeutic targets.

Integrating machine learning techniques into robust data enrichment approach and its application to gene expression data.

The availability of enough samples for effective analysis and knowledge discovery has been a challenge in the research community, especially in the area of gene expression data analysis. Thus, the approaches being developed for data analysis have mostly suffered from the lack of enough data to train and test the constructed models. We argue that the process of sample generation could be successfully automated by employing some sophisticated machine learning techniques. An automated sample generation framework could successfully complement the actual sample generation from real cases. This argument is validated in this paper by describing a framework that integrates multiple models (perspectives) for sample generation. We illustrate its applicability for producing new gene expression data samples, a highly demanding area that has not received attention. The three perspectives employed in the process are based on models that are not closely related. The independence eliminates the bias of having the produced approach covering only certain characteristics of the domain and leading to samples skewed towards one direction. The first model is based on the Probabilistic Boolean Network (PBN) representation of the gene regulatory network underlying the given gene expression data. The second model integrates Hierarchical Markov Model (HIMM) and the third model employs a genetic algorithm in the process. Each model learns as much as possible characteristics of the domain being analysed and tries to incorporate the learned characteristics in generating new samples. In other words, the models base their analysis on domain knowledge implicitly present in the data itself. The developed framework has been extensively tested by checking how the new samples complement the original samples. The produced results are very promising in showing the effectiveness, usefulness and applicability of the proposed multi-model framework.

IDH mutation detection in formalin-fixed paraffin-embedded gliomas using multiplex PCR and single-base extension.

Isocitrate dehydrogenase (IDH) genes are mutated in a significant portion of gliomas, myeloid leukemias and chondroid neoplasms. In gliomas, IDH mutations are prognostic, as those tumors with the mutation are associated with a proneural subclass and have longer survival compared with those without the mutation. We developed a simple, PCR-based SNaPshot® assay (Life Technologies, Carlsbad, CA, USA) to detect IDH1/2 mutations. This protocol combines a single, multiplexed PCR reaction using gene specific primers followed by a single, multiplexed SNaPshot reaction and detection by capillary electrophoresis. In a blinded study of 32 paraffin-embedded glioma specimens previously screened for IDH mutations by a PCR/direct sequencing method, concordance of our IDH SNaPshot test with sequencing was 100%. We performed the assay on an additional 57 specimens submitted for diagnostic IDH mutation evaluation. Data analysis was much faster and easier to perform than analysis of the sequencing data, and results could be obtained in 1 day from DNA extraction to analysis. Furthermore, we could readily identify a mixture of 5% mutant allele vs. 95% wild-type allele in our SNaPshot assay, in comparison to approximately 20% mutant allele in our PCR-sequencing assay. Our assay represents a fast, sensitive, straightforward method of reliably detecting common mutations of IDH genes in glial neoplasms, or other tumors.

A closer look at "social" boundary genes reveals knowledge to gene expression profiles.

As social network analysis is gaining popularity in modeling real world problems, the task of applying the social network model concepts and notions to biological data is still one of the most attractive research problems to be addressed. According, our work described in this paper focuses on a particular set of genes that reside on the community boundaries in gene co-expression networks. Stemmed from community mining problem in social networks, peripheries of communities (i.e., boundaries) can be used to aid certain biological analysis. The proposed method consists of three parts: 1) Finding communities of gene co-expression networks through clustering. 2) Analyzing stability of community structures by Monte Carlo method. 3) Designing of dynamic adoption of boundaries using geometric convexity. We validated our findings using breast cancer gene expression data from various studies. Our approach contributes to the new branch of applying social network mechanisms in biological data analysis, leading to new data mining strategies implied by witnessing social behaviors in gene expression analysis.

AKT loss in human epithelial cells treated with severe hypoxia.

Cancer cells which can survive and or proliferate in hypoxia may be resistant to anti-cancer treatment. In our previous work, we showed that we could group cell lines treated with severe hypoxia into either hypoxia-induced cell cycle arrest-sensitive or resistant phenotypes, and hypoxia-induced cell death (HCD)-sensitive or resistant phenotypes. We showed that the resistant phenotypes were associated with high levels of active-AKT in late hypoxia and sensitive cells were associated with decreased or undetectable levels of AKT in late hypoxia. We have now extended our findings to numerous other cell lines. We show that HCD and loss of AKT is cell density dependent, and both AKT1 and AKT2 isoforms are lost in late hypoxia. Loss of AKT is most likely due to regulated degradation, as transcription of AKT isoforms is unchanged in hypoxia, and AKT is not significantly translocated to the nucleus to account for its disappearance from cytoplasmic lysates. Interestingly, inhibitors of proteosome, calpain or caspase-mediated proteolysis did not significantly block AKT loss. Inhibition of autophagy using diverse lysosome-targeted autophagy inhibitors also did not block AKT loss, however autophagy inhibitors which block general PI3K activity, such as 3-methyladenine or LY294002, were effective inhibitors of AKT loss in late hypoxia. Interestingly, those inhibitors also blocked HCD in an HCD-sensitive cancer cell line. Inhibitors of proteolytic pathways which did not block AKT loss also did not block HCD in HeLa. Our investigations support a model by which AKT is a major switch involved in regulating hypoxia-induced cell death.

Adaptive multi-agent architecture for functional sequence motifs recognition.

MOTIVATION: Accurate genome annotation or protein function prediction requires precise recognition of functional sequence motifs. Many computational motif prediction models have been proposed. Due to the complexity of the biological data, it may be desirable to apply an integrated approach that uses multiple models for analysis. RESULTS: In this article, we propose a novel multi-agent architecture for the general purpose of functional sequence motif recognition. The approach takes advantage of the synergy provided by multiple agents through the employment of different agents equipped with distinctive problem solving skills and promotes the collaborations among them through decision maker (DM) agents that work as classifier ensembles. A genetic algorithm-based fusion strategy is applied which offers evolutionary property to the DM agents. The consistency and robustness of the system are maintained by an evolvable agent that mediates the team of the ensemble agents. The combined effort of a recommendation system (Seer) and the self-learning mediator agent yields a successful identification of the most efficient agent deployment scheme at an early stage of the experimentation process, which has the potential of greatly reducing the computational cost of the system. Two concrete systems are constructed that aim at predicting two important sequence motifs-the translational initiation sites (TISs) and the core promoters. With the incorporation of three distinctive problem solver agents, the TIS predictor consistently outperforms most of the state-of-the-art approaches under investigation. Integrating three existing promoter predictors, our system is able to yield consistently good performance. AVAILABILITY: The program (MotifMAS) and the datasets are available upon request.

Representative transcript sets for evaluating a translational initiation sites predictor.

BACKGROUND: Translational initiation site (TIS) prediction is a very important and actively studied topic in bioinformatics. In order to complete a comparative analysis, it is desirable to have several benchmark data sets which can be used to test the effectiveness of different algorithms. An ideal benchmark data set should be reliable, representative and readily available. Preferably, proteins encoded by members of the data set should also be representative of the protein population actually expressed in cellular specimens. RESULTS: In this paper, we report a general algorithm for constructing a reliable sequence collection that only includes mRNA sequences whose corresponding protein products present an average profile of the general protein population of a given organism, with respect to three major structural parameters. Four representative transcript collections, each derived from a model organism, have been obtained following the algorithm we propose. Evaluation of these data sets shows that they are reasonable representations of the spectrum of proteins obtained from cellular proteomic studies. Six state-of-the-art predictors have been used to test the usefulness of the construction algorithm that we proposed. Comparative study which reports the predictors' performance on our data set as well as three other existing benchmark collections has demonstrated the actual merits of our data sets as benchmark testing collections. CONCLUSION: The proposed data set construction algorithm has demonstrated its property of being a general and widely applicable scheme. Our comparison with published proteomic studies has shown that the expression of our data set of transcripts generates a polypeptide population that is representative of that obtained from evaluation of biological specimens. Our data set thus represents "real world" transcripts that will allow more accurate evaluation of algorithms dedicated to identification of TISs, as well as other translational regulatory motifs within mRNA sequences. The algorithm proposed by us aims at compiling a redundancy-free data set by removing redundant copies of homologous proteins. The existence of such data sets may be useful for conducting statistical analyses of protein sequence-structure relations. At the current stage, our approach's focus is to obtain an "average" protein data set for any particular organism without posing much selection bias. However, with the three major protein structural parameters deeply integrated into the scheme, it would be a trivial task to extend the current method for obtaining a more selective protein data set, which may facilitate the study of some particular protein structure.