Effect of the LSD1 inhibitor RN-1 on γ-globin and global gene expression during erythroid differentiation in baboons (Papio anubis).

Elevated levels of Fetal Hemoglobin interfere with polymerization of sickle hemoglobin thereby reducing anemia, lessening the severity of symptoms, and increasing life span of patients with sickle cell disease. An affordable, small molecule drug that stimulates HbF expression in vivo would be ideally suited to treat the large numbers of SCD patients that exist worldwide. Our previous work showed that administration of the LSD1 (KDM1A) inhibitor RN-1 to normal baboons increased Fetal Hemoglobin (HbF) and was tolerated over a prolonged treatment period. HbF elevations were associated with changes in epigenetic modifications that included increased levels of H3K4 di-and tri-methyl lysine at the γ-globin promoter. While dramatic effects of the loss of LSD1 on hematopoietic differentiation have been observed in murine LSD1 gene deletion and silencing models, the effect of pharmacological inhibition of LSD1 in vivo on hematopoietic differentiation is unknown. The goal of these experiments was to investigate the in vivo mechanism of action of the LSD1 inhibitor RN-1 by determining its effect on γ-globin expression in highly purified subpopulations of bone marrow erythroid cells enriched for varying stages of erythroid differentiation isolated directly from baboons treated with RN-1 and also by investigating the effect of RN1 on the global transcriptome in a highly purified population of proerythroblasts. Our results show that RN-1 administered to baboons targets an early event during erythroid differentiation responsible for γ-globin repression and increases the expression of a limited number of genes including genes involved in erythroid differentiation such as GATA2, GFi-1B, and LYN.

Wnt pathway inhibitors are upregulated in XLH dental pulp cells in response to odontogenic differentiation.

X-linked hypophosphatemia (XLH) represents the most common form of familial hypophosphatemia. Although significant advances have been made in the treatment of bone pathology, patients undergoing therapy continue to experience significantly decreased oral health-related quality of life. The following study addresses this persistent oral disease by further investigating the effect of DMP1 expression on the differentiation of XLH dental pulp cells. Dental pulp cells were isolated from the third molars of XLH and healthy controls and stable transduction of full-length human DMP1 were achieved. RNA sequencing was performed to evaluate the genetic changes following the induction of odontogenic differentiation. RNAseq data shows the upregulation of inhibitors of the canonical Wnt pathway in XLH cells, while constitutive expression of full-length DMP1 in XLH cells reversed this effect during odontogenic differentiation. These results imply that inhibition of the canonical Wnt pathway may contribute to the pathophysiology of XLH and suggest a new therapeutic strategy for the management of oral disease.

Dysregulated microRNA expression in IL-4 transgenic mice, an animal model of atopic dermatitis.

IL-4 plays an important role in the pathogenesis of atopic dermatitis (AD). Previously we showed that the expression of genes in chemotaxis, angiogenesis, inflammation and barrier functions is dysregulated in IL-4 transgenic (Tg) mice, a well-characterized AD mouse model. In this study, we aim to study differential expression of microRNAs in IL-4 Tg mice. As compared with wild-type mice, we found that 10 and 79 microRNAs are dysregulated in the skin of IL-4 mice before and after the onset of skin lesions, respectively. Bioinformatic analysis and previous reports show that these dysregulated microRNAs may be involved in the NF-κB, TLRs, IL-4/IL-13, MAPK and other pathways. We also found that miR-139-5p and miR-196b-3p are significantly up-regulated in the peripheral blood of IL-4 Tg mice. Taken together, our data have identified many dysregulated microRNAs in IL-4 Tg mice, which may play important roles in AD pathogenesis and pathophysiology.

Longitudinal Analysis of the T-cell Receptor Repertoire in Graft-infiltrating Lymphocytes Following Hand Transplantation.

BACKGROUND: T lymphocyte-mediated acute rejection is a significant complication following solid organ transplantation. Standard methods of monitoring for acute rejection rely on assessing histological tissue damage but do not define the immunopathogenesis. Additionally, current therapies for rejection broadly blunt cellular immunity, creating a high risk for opportunistic infections. There is, therefore, a need to better understand the process of acute cellular rejection to help develop improved prognostic tests and narrowly targeted therapies. METHODS: Through next-generation sequencing, we characterized and compared the clonal T-cell receptor (TCR) repertoires of graft-infiltrating lymphocytes (GILs) and blood-derived lymphocytes from a hand transplant recipient over 420 days following transplantation. We also tracked the TCR clonal persistence and V beta (BV) gene usage, evaluating overlap between these 2 compartments. RESULTS: TCR repertoires of blood and GIL populations remained distinct throughout the sampling period, and differential BV usage was consistently seen between these compartments. GIL TCR clones persisted over time and were seen in only limited frequency in the blood T-lymphocyte populations. CONCLUSIONS: We demonstrate that blood monitoring of TCR clones does not reveal the pathogenic process of acute cellular rejection in transplanted tissue. GILs show clonal persistence with biased BV usage, suggesting that tissue TCR clonal monitoring could be useful, although a deeper understanding is necessary to prognosticate rejection based on TCR clonal repertoires. Finally, the distinct TCR BV usage bias in GILs raises the possibility for prevention and therapy of acute cellular rejection based on targeting of specific TCR clones.

HDAC inhibitors elicit metabolic reprogramming by targeting super-enhancers in glioblastoma models.

The Warburg effect is a tumor-related phenomenon that could potentially be targeted therapeutically. Here, we showed that glioblastoma (GBM) cultures and patients' tumors harbored super-enhancers in several genes related to the Warburg effect. By conducting a transcriptome analysis followed by ChIP-Seq coupled with a comprehensive metabolite analysis in GBM models, we found that FDA-approved global (panobinostat, vorinostat) and selective (romidepsin) histone deacetylase (HDAC) inhibitors elicited metabolic reprogramming in concert with disruption of several Warburg effect-related super-enhancers. Extracellular flux and carbon-tracing analyses revealed that HDAC inhibitors blunted glycolysis in a c-Myc-dependent manner and lowered ATP levels. This resulted in the engagement of oxidative phosphorylation (OXPHOS) driven by elevated fatty acid oxidation (FAO), rendering GBM cells dependent on these pathways. Mechanistically, interference with HDAC1/-2 elicited a suppression of c-Myc protein levels and a concomitant increase in 2 transcriptional drivers of oxidative metabolism, PGC1α and PPARD, suggesting an inverse relationship. Rescue and ChIP experiments indicated that c-Myc bound to the promoter regions of PGC1α and PPARD to counteract their upregulation driven by HDAC1/-2 inhibition. Finally, we demonstrated that combination treatment with HDAC and FAO inhibitors extended animal survival in patient-derived xenograft model systems in vivo more potently than single treatments in the absence of toxicity.

Complete Genome Sequence of a USA100 Methicillin-Resistant Staphylococcus aureus Strain.

We report the genome sequence of a methicillin-resistant Staphylococcus aureus (MRSA) strain, isolated from a surgical intensive care unit. This completely closed genome of a USA100 isolate contains a major chromosome and a plasmid and will serve as a reference genome for genetic analysis of MRSA strains.

Expression profiling of genes regulated by sphingosine kinase1 signaling in a murine model of hyperoxia induced neonatal bronchopulmonary dysplasia.

BACKGROUND: Sphingosine- 1-Phosphate (S1P) is a bioactive lipid and an intracellular as well as an extracellular signaling molecule. S1P ligand specifically binds to five related cell surface G-protein-coupled receptors (S1P1-5). S1P levels are tightly regulated by its synthesis catalyzed by sphingosine kinases (SphKs) 1 & 2 and catabolism by S1P phosphatases, lipid phosphate phosphatases and S1P lyase. We previously reported that knock down of SphK1 (Sphk1 -/- ) in a neonatal mouse BPD model conferred significant protection against hyperoxia induced lung injury. To better understand the underlying molecular mechanisms, genome-wide gene expression profiling was performed on mouse lung tissue using Affymetrix MoGene 2.0 array. RESULTS: Two-way ANOVA analysis was performed and differentially expressed genes under hyperoxia were identified using Sphk1 -/- mice and their wild type (WT) equivalents. Pathway (PW) enrichment analyses identified several signaling pathways that are likely to play a key role in hyperoxia induced lung injury in the neonates. These included signaling pathways that were anticipated such as those involved in lipid signaling, cell cycle regulation, DNA damage/apoptosis, inflammation/immune response, and cell adhesion/extracellular matrix (ECM) remodeling. We noted hyperoxia induced downregulation of the expression of genes related to mitotic spindle formation in the WT which was not observed in Sphk1 -/- neonates. Our data clearly suggests a role for SphK1 in neonatal hyperoxic lung injury through elevated inflammation and apoptosis in lung tissue. Further, validation by RT-PCR on 24 differentially expressed genes showed 83% concordance both in terms of fold change and vectorial changes. Our findings are in agreement with previously reported human BPD microarray data and completely support our published in vivo findings. In addition, the data also revealed a significant role for additional unanticipitated signaling pathways involving Wnt and GADD45. CONCLUSION: Using SphK1 knockout mice and differential gene expression analysis, we have shown here that S1P/SphK1 signaling plays a key role in promoting hyperoxia induced DNA damage, inflammation, apoptosis and ECM remodeling in neonatal lungs. It also appears to suppress pro-survival cellular responses involved in normal lung development. We therefore propose SphK1 as a therapeutic target for the development drugs to combat BPD.

Investigation of Respiratory Syncytial Virus Outbreak on an Adult Stem Cell Transplant Unit by Use of Whole-Genome Sequencing.

A viral whole-genome sequencing (WGS) strategy, based on PCR amplification followed by next-generation sequencing, was used to investigate a nosocomial respiratory syncytial virus-B (RSV-B) outbreak in a hematology-oncology and stem cell transplant unit. RSV-B genomes from 16 patients and health care workers (HCWs) suspected to be involved in the outbreak were compared to RSV-B genomes that were acquired from outpatients during the same time period but epidemiologically unrelated to the outbreak. Phylogenetic analysis of the whole genome identified a cluster of 11 patients and HCWs who had an identical RSV-B strain which was clearly distinct from strains recovered from individuals unrelated to the outbreak. Sequence variation of the glycoprotein (G) gene alone was insufficient to distinguish the outbreak strains from the outbreak-unrelated strains, thereby demonstrating that WGS is valuable for local outbreak investigation.

Association of circulating transcriptomic profiles with mortality in sickle cell disease.

Sickle cell disease (SCD) complications are associated with increased morbidity and risk of mortality. We sought to identify a circulating transcriptomic profile predictive of these poor outcomes in SCD. Training and testing cohorts consisting of adult patients with SCD were recruited and prospectively followed. A pathway-based signature derived from grouping peripheral blood mononuclear cell transcriptomes distinguished 2 patient clusters with differences in survival in the training cohort. These findings were validated in a testing cohort in which the association between cluster 1 molecular profiling and mortality remained significant in a fully adjusted model. In a third cohort of West African children with SCD, cluster 1 differentiated SCD severity using a published scoring index. Finally, a risk score composed of assigning weights to cluster 1 profiling, along with established clinical risk factors using tricuspid regurgitation velocity, white blood cell count, history of acute chest syndrome, and hemoglobin levels, demonstrated a higher hazard ratio for mortality in both the training and testing cohorts compared with clinical risk factors or cluster 1 data alone. Circulating transcriptomic profiles are a powerful method to risk-stratify severity of disease and poor outcomes in both children and adults, respectively, with SCD and highlight potential associated molecular pathways.

Wnt proteins synergize to activate ß-catenin signaling.

Wnt ligands are involved in diverse signaling pathways that are active during development, maintenance of tissue homeostasis and in various disease states. While signaling regulated by individual Wnts has been extensively studied, Wnts are rarely expressed alone, and the consequences of Wnt gene co-expression are not well understood. Here, we studied the effect of co-expression of Wnts on the ß-catenin signaling pathway. While some Wnts are deemed 'non-canonical' due to their limited ability to activate ß-catenin when expressed alone, unexpectedly, we find that multiple Wnt combinations can synergistically activate ß-catenin signaling in multiple cell types. WNT1- and WNT7B-mediated synergistic Wnt signaling requires FZD5, FZD8 and LRP6, as well as the WNT7B co-receptors GPR124 (also known as ADGRA2) and RECK. Unexpectedly, this synergistic signaling occurs downstream of ß-catenin stabilization, and is correlated with increased lysine acetylation of ß-catenin. Wnt synergy provides a general mechanism to confer increased combinatorial control over this important regulatory pathway.

Differentially Expressed miRNAs in Hepatocellular Carcinoma Target Genes in the Genetic Information Processing and Metabolism Pathways.

To date, studies of the roles of microRNAs (miRNAs) in hepatocellular carcinoma (HCC) have either focused on specific individual miRNAs and a small number of suspected targets or simply reported a list of differentially expressed miRNAs based on expression profiling. Here, we seek a more in-depth understanding of the roles of miRNAs and their targets in HCC by integrating the miRNA and messenger RNA (mRNA) expression profiles of tumorous and adjacent non-tumorous liver tissues of 100 HCC patients. We assessed the levels of 829 mature miRNAs, of which 32 were significantly differentially expressed. Statistical analysis indicates that six of these miRNAs regulate a significant proportion of their in silico predicted target mRNAs. Three of these miRNAs (miR-26a, miR-122, and miR-130a) were down-regulated in HCC, and their up-regulated gene targets are primarily associated with aberrant cell proliferation that involves DNA replication, transcription and nucleotide metabolism. The other three miRNAs (miR-21, miR-93, and miR-221) were up-regulated in HCC, and their down-regulated gene targets are primarily involved in metabolism and immune system processes. We further found evidence for a coordinated miRNA-induced regulation of important cellular processes, a finding to be considered when designing therapeutic applications based on miRNAs.

Abundant copy-number loss of CYCLOPS and STOP genes in gastric adenocarcinoma.

BACKGROUND: Gastric cancer, a leading cause of cancer death worldwide, has been little studied compared with other cancers that impose similar health burdens. Our goal is to assess genomic copy-number loss and the possible functional consequences and therapeutic implications thereof across a large series of gastric adenocarcinomas. METHODS: We used high-density single-nucleotide polymorphism microarrays to determine patterns of copy-number loss and allelic imbalance in 74 gastric adenocarcinomas. We investigated whether suppressor of tumorigenesis and/or proliferation (STOP) genes are associated with genomic copy-number loss. We also analyzed the extent to which copy-number loss affects Copy-number alterations Yielding Cancer Liabilities Owing to Partial losS (CYCLOPS) genes-genes that may be attractive targets for therapeutic inhibition when partially deleted. RESULTS: The proportion of the genome subject to copy-number loss varies considerably from tumor to tumor, with a median of 5.5 %, and a mean of 12 % (range 0-58.5 %). On average, 91 STOP genes were subject to copy-number loss per tumor (median 35, range 0-452), and STOP genes tended to have lower copy-number compared with the rest of the genes. Furthermore, on average, 1.6 CYCLOPS genes per tumor were both subject to copy-number loss and downregulated, and 51.4 % of the tumors had at least one such gene. CONCLUSIONS: The enrichment of STOP genes in regions of copy-number loss indicates that their deletion may contribute to gastric carcinogenesis. Furthermore, the presence of several deleted and downregulated CYCLOPS genes in some tumors suggests potential therapeutic targets in these tumors.

Integrative analysis of head and neck cancer identifies two biologically distinct HPV and three non-HPV subtypes.

PURPOSE: Current classification of head and neck squamous cell carcinomas (HNSCC) based on anatomic site and stage fails to capture biologic heterogeneity or adequately inform treatment. EXPERIMENTAL DESIGN: Here, we use gene expression-based consensus clustering, copy number profiling, and human papillomavirus (HPV) status on a clinically homogenous cohort of 134 locoregionally advanced HNSCCs with 44% HPV(+) tumors together with additional cohorts, which in total comprise 938 tumors, to identify HNSCC subtypes and discover several subtype-specific, translationally relevant characteristics. RESULTS: We identified five subtypes of HNSCC, including two biologically distinct HPV subtypes. One HPV(+) and one HPV(-) subtype show a prominent immune and mesenchymal phenotype. Prominent tumor infiltration with CD8(+) lymphocytes characterizes this inflamed/mesenchymal subtype, independent of HPV status. Compared with other subtypes, the two HPV subtypes show low expression and no copy number events for EGFR/HER ligands. In contrast, the basal subtype is uniquely characterized by a prominent EGFR/HER signaling phenotype, negative HPV-status, as well as strong hypoxic differentiation not seen in other subtypes. CONCLUSION: Our five-subtype classification provides a comprehensive overview of HPV(+) as well as HPV(-) HNSCC biology with significant translational implications for biomarker development and personalized care for patients with HNSCC.

Lipidomics identifies a requirement for peroxisomal function during influenza virus replication.

Influenza virus acquires a host-derived lipid envelope during budding, yet a convergent view on the role of host lipid metabolism during infection is lacking. Using a mass spectrometry-based lipidomics approach, we provide a systems-scale perspective on membrane lipid dynamics of infected human lung epithelial cells and purified influenza virions. We reveal enrichment of the minor peroxisome-derived ether-linked phosphatidylcholines relative to bulk ester-linked phosphatidylcholines in virions as a unique pathogenicity-dependent signature for influenza not found in other enveloped viruses. Strikingly, pharmacological and genetic interference with peroxisomal and ether lipid metabolism impaired influenza virus production. Further integration of our lipidomics results with published genomics and proteomics data corroborated altered peroxisomal lipid metabolism as a hallmark of influenza virus infection in vitro and in vivo. Influenza virus may therefore tailor peroxisomal and particularly ether lipid metabolism for efficient replication.

ARTS: automated randomization of multiple traits for study design.

UNLABELLED: Collecting data from large studies on high-throughput platforms, such as microarray or next-generation sequencing, typically requires processing samples in batches. There are often systematic but unpredictable biases from batch-to-batch, so proper randomization of biologically relevant traits across batches is crucial for distinguishing true biological differences from experimental artifacts. When a large number of traits are biologically relevant, as is common for clinical studies of patients with varying sex, age, genotype and medical background, proper randomization can be extremely difficult to prepare by hand, especially because traits may affect biological inferences, such as differential expression, in a combinatorial manner. Here we present ARTS (automated randomization of multiple traits for study design), which aids researchers in study design by automatically optimizing batch assignment for any number of samples, any number of traits and any batch size. AVAILABILITY AND IMPLEMENTATION: ARTS is implemented in Perl and is available at github.com/mmaiensc/ARTS. ARTS is also available in the Galaxy Tool Shed, and can be used at the Galaxy installation hosted by the UIC Center for Research Informatics (CRI) at galaxy.cri.uic.edu.

Identification of molecular subtypes of gastric cancer with different responses to PI3-kinase inhibitors and 5-fluorouracil.

BACKGROUND & AIMS: Almost all gastric cancers are adenocarcinomas, which have considerable heterogeneity among patients. We sought to identify subtypes of gastric adenocarcinomas with particular biological properties and responses to chemotherapy and targeted agents. METHODS: We compared gene expression patterns among 248 gastric tumors; using a robust method of unsupervised clustering, consensus hierarchical clustering with iterative feature selection, we identified 3 major subtypes. We developed a classifier for these subtypes and validated it in 70 tumors from a different population. We identified distinct genomic and epigenomic properties of the subtypes. We determined drug sensitivities of the subtypes in primary tumors using clinical survival data, and in cell lines through high-throughput drug screening. RESULTS: We identified 3 subtypes of gastric adenocarcinoma: proliferative, metabolic, and mesenchymal. Tumors of the proliferative subtype had high levels of genomic instability, TP53 mutations, and DNA hypomethylation. Cancer cells of the metabolic subtype were more sensitive to 5-fluorouracil than the other subtypes. Furthermore, in 2 independent groups of patients, those with tumors of the metabolic subtype appeared to have greater benefits with 5-fluorouracil treatment. Tumors of the mesenchymal subtype contain cells with features of cancer stem cells, and cell lines of this subtype are particularly sensitive to phosphatidylinositol 3-kinase-AKT-mTOR inhibitors in vitro. CONCLUSIONS: Based on gene expression patterns, we classified gastric cancers into 3 subtypes, and validated these in an independent set of tumors. The subgroups have differences in molecular and genetic features and response to therapy; this information might be used to select specific treatment approaches for patients with gastric cancer.

mTORC1 inhibition restricts inflammation-associated gastrointestinal tumorigenesis in mice.

Gastrointestinal cancers are frequently associated with chronic inflammation and excessive secretion of IL-6 family cytokines, which promote tumorigenesis through persistent activation of the GP130/JAK/STAT3 pathway. Although tumor progression can be prevented by genetic ablation of Stat3 in mice, this transcription factor remains a challenging therapeutic target with a paucity of clinically approved inhibitors. Here, we uncovered parallel and excessive activation of mTOR complex 1 (mTORC1) alongside STAT3 in human intestinal-type gastric cancers (IGCs). Furthermore, in a preclinical mouse model of IGC, GP130 ligand administration simultaneously activated mTORC1/S6 kinase and STAT3 signaling. We therefore investigated whether mTORC1 activation was required for inflammation-associated gastrointestinal tumorigenesis. Strikingly, the mTORC1-specific inhibitor RAD001 potently suppressed initiation and progression of both murine IGC and colitis-associated colon cancer. The therapeutic effect of RAD001 was associated with reduced tumor vascularization and cell proliferation but occurred independently of STAT3 activity. We analyzed the mechanism of GP130-mediated mTORC1 activation in cells and mice and revealed a requirement for JAK and PI3K activity but not for GP130 tyrosine phosphorylation or STAT3. Our results suggest that GP130-dependent activation of the druggable PI3K/mTORC1 pathway is required for inflammation-associated gastrointestinal tumorigenesis. These findings advocate clinical application of PI3K/mTORC1 inhibitors for the treatment of corresponding human malignancies.

A comprehensive survey of genomic alterations in gastric cancer reveals systematic patterns of molecular exclusivity and co-occurrence among distinct therapeutic targets.

OBJECTIVE: Gastric cancer is a major gastrointestinal malignancy for which targeted therapies are emerging as treatment options. This study sought to identify the most prevalent molecular targets in gastric cancer and to elucidate systematic patterns of exclusivity and co-occurrence among these targets, through comprehensive genomic analysis of a large panel of gastric cancers. DESIGN: Using high-resolution single nucleotide polymorphism arrays, copy number alterations were profiled in a panel of 233 gastric cancers (193 primary tumours, 40 cell lines) and 98 primary matched gastric non-malignant samples. For selected alterations, their impact on gene expression and clinical outcome were evaluated. RESULTS: 22 recurrent focal alterations (13 amplifications and nine deletions) were identified. These included both known targets (FGFR2, ERBB2) and also novel genes in gastric cancer (KLF5, GATA6). Receptor tyrosine kinase (RTK)/RAS alterations were found to be frequent in gastric cancer. This study also demonstrates, for the first time, that these alterations occur in a mutually exclusive fashion, with KRAS gene amplifications highlighting a clinically relevant but previously underappreciated gastric cancer subgroup. FGFR2-amplified gastric cancers were also shown to be sensitive to dovitinib, an orally bioavailable FGFR/VEGFR targeting agent, potentially representing a subtype-specific therapy for FGFR2-amplified gastric cancers. CONCLUSION: The study demonstrates the existence of five distinct gastric cancer patient subgroups, defined by the signature genomic alterations FGFR2 (9% of tumours), KRAS (9%), EGFR (8%), ERBB2 (7%) and MET (4%). Collectively, these subgroups suggest that at least 37% of gastric cancer patients may be potentially treatable by RTK/RAS directed therapies.

Assessing protein similarity with Gene Ontology and its use in subnuclear localization prediction.

BACKGROUND: The accomplishment of the various genome sequencing projects resulted in accumulation of massive amount of gene sequence information. This calls for a large-scale computational method for predicting protein localization from sequence. The protein localization can provide valuable information about its molecular function, as well as the biological pathway in which it participates. The prediction of localization of a protein at subnuclear level is a challenging task. In our previous work we proposed an SVM-based system using protein sequence information for this prediction task. In this work, we assess protein similarity with Gene Ontology (GO) and then improve the performance of the system by adding a module of nearest neighbor classifier using a similarity measure derived from the GO annotation terms for protein sequences. RESULTS: The performance of the new system proposed here was compared with our previous system using a set of proteins resided within 6 localizations collected from the Nuclear Protein Database (NPD). The overall MCC (accuracy) is elevated from 0.284 (50.0%) to 0.519 (66.5%) for single-localization proteins in leave-one-out cross-validation; and from 0.420 (65.2%) to 0.541 (65.2%) for an independent set of multi-localization proteins. The new system is available at http://array.bioengr.uic.edu/subnuclear.htm. CONCLUSION: The prediction of protein subnuclear localizations can be largely influenced by various definitions of similarity for a pair of proteins based on different similarity measures of GO terms. Using the sum of similarity scores over the matched GO term pairs for two proteins as the similarity definition produced the best predictive outcome. Substantial improvement in predicting protein subnuclear localizations has been achieved by combining Gene Ontology with sequence information.

An SVM-based system for predicting protein subnuclear localizations.

BACKGROUND: The large gap between the number of protein sequences in databases and the number of functionally characterized proteins calls for the development of a fast computational tool for the prediction of subnuclear and subcellular localizations generally applicable to protein sequences. The information on localization may reveal the molecular function of novel proteins, in addition to providing insight on the biological pathways in which they function. The bulk of past work has been focused on protein subcellular localizations. Furthermore, no specific tool has been dedicated to prediction at the subnuclear level, despite its high importance. In order to design a suitable predictive system, the extraction of subtle sequence signals that can discriminate among proteins with different subnuclear localizations is the key. RESULTS: New kernel functions used in a support vector machine (SVM) learning model are introduced for the measurement of sequence similarity. The k-peptide vectors are first mapped by a matrix of high-scored pairs of k-peptides which are measured by BLOSUM62 scores. The kernels, measuring the similarity for sequences, are then defined on the mapped vectors. By combining these new encoding methods, a multi-class classification system for the prediction of protein subnuclear localizations is established for the first time. The performance of the system is evaluated with a set of proteins collected in the Nuclear Protein Database (NPD). The overall accuracy of prediction for 6 localizations is about 50% (vs. random prediction 16.7%) for single localization proteins in the leave-one-out cross-validation; and 65% for an independent set of multi-localization proteins. This integrated system can be accessed at http://array.bioengr.uic.edu/subnuclear.htm. CONCLUSION: The integrated system benefits from the combination of predictions from several SVMs based on selected encoding methods. Finally, the predictive power of the system is expected to improve as more proteins with known subnuclear localizations become available.