Segment and fit thresholding: a new method for image analysis applied to microarray and immunofluorescence data.

Experiments involving the high-throughput quantification of image data require algorithms for automation. A challenge in the development of such algorithms is to properly interpret signals over a broad range of image characteristics, without the need for manual adjustment of parameters. Here we present a new approach for locating signals in image data, called Segment and Fit Thresholding (SFT). The method assesses statistical characteristics of small segments of the image and determines the best-fit trends between the statistics. Based on the relationships, SFT identifies segments belonging to background regions; analyzes the background to determine optimal thresholds; and analyzes all segments to identify signal pixels. We optimized the initial settings for locating background and signal in antibody microarray and immunofluorescence data and found that SFT performed well over multiple, diverse image characteristics without readjustment of settings. When used for the automated analysis of multicolor, tissue-microarray images, SFT correctly found the overlap of markers with known subcellular localization, and it performed better than a fixed threshold and Otsu's method for selected images. SFT promises to advance the goal of full automation in image analysis.

Anoikis-resistant subpopulations of human osteosarcoma display significant chemoresistance and are sensitive to targeted epigenetic therapies predicted by expression profiling.

BACKGROUND: Osteosarcoma (OS) is the most common type of solid bone cancer, with latent metastasis being a typical mode of disease progression and a major contributor to poor prognosis. For this to occur, cells must resist anoikis and be able to recapitulate tumorigenesis in a foreign microenvironment. Finding novel approaches to treat osteosarcoma and target those cell subpopulations that possess the ability to resist anoikis and contribute to metastatic disease is imperative. Here we investigate anchorage-independent (AI) cell growth as a model to better characterize anoikis resistance in human osteosarcoma while using an expression profiling approach to identify and test targetable signaling pathways. METHODS: Established human OS cell lines and patient-derived human OS cell isolates were subjected to growth in either adherent or AI conditions using Ultra-Low Attachment plates in identical media conditions. Growth rate was assessed using cell doubling times and chemoresistance was assessed by determining cell viability in response to a serial dilution of either doxorubicin or cisplatin. Gene expression differences were examined using quantitative reverse-transcription PCR and microarray with principal component and pathway analysis. In-vivo OS xenografts were generated by either subcutaneous or intratibial injection of adherent or AI human OS cells into athymic nude mice. Statistical significance was determined using student's t-tests with significance set at α=0.05. RESULTS: We show that AI growth results in a global gene expression profile change accompanied by significant chemoresistance (up to 75 fold, p<0.05). AI cells demonstrate alteration of key mediators of mesenchymal differentiation (ß-catenin, Runx2), stemness (Sox2), proliferation (c-myc, Akt), and epigenetic regulation (HDAC class 1). AI cells were equally tumorigenic as their adherent counterparts, but showed a significantly decreased rate of growth in-vitro and in-vivo (p<0.05). Treatment with the pan-histone deacetylase inhibitor vorinostat and the DNA methyltransferase inhibitor 5-azacytidine mitigated AI growth, while 5-azacytidine sensitized anoikis-resistant cells to doxorubicin (p<0.05). CONCLUSIONS: These data demonstrate remarkable plasticity in anoikis-resistant human osteosarcoma subpopulations accompanied by a rapid development of chemoresistance and altered growth rates mirroring the early stages of latent metastasis. Targeting epigenetic regulation of this process may be a viable therapeutic strategy.

Using a rhabdomyosarcoma patient-derived xenograft to examine precision medicine approaches and model acquired resistance.

BACKGROUND: Precision (Personalized) medicine has the potential to revolutionize patient health care especially for many cancers where the fundamental disease etiology remains either elusive or has no available therapy. Here we outline a study in alveolar rhabdomyosarcoma, in which we use gene expression profiling and a series of drug prediction algorithms combined with a matched patient-derived xenograft (PDX) model to test bioinformatically predicted therapies. PROCEDURE: A PDX model was developed from a patient biopsy and a number of drugs identified using gene expression analysis in combination with drug prediction algorithms. Drugs chosen from each of the predictive methodologies, along with the patient's standard-of-care therapy (ICE-T), were tested in vivo in the PDX tumor. A second study was initiated using the tumors that re-grew following the ICE-T treatment. Further expression analysis identified additional therapies with potential anti-tumor efficacy. RESULTS: A number of the predicted therapies were found to be active against the tumors in particular BGJ398 (FGFR2) and ICE-T. Re-transplanted ICE-T treated tumorgrafts demonstrated a decreased response to ICE-T recapitulating the patient's refractory disease. Gene expression profiling of the ICE-T treated tumorgrafts identified cytarabine (SLC29A1) as a potential therapy, which was shown, along with BGJ398, to be highly active in vivo. CONCLUSIONS: This study illustrates that PDX models are suitable surrogates for testing potential therapeutic strategies based on gene expression analysis, modeling clinical drug resistance and hold the potential to assist in guiding prospective patient care.

Molecular-guided therapy predictions reveal drug resistance phenotypes and treatment alternatives in malignant peripheral nerve sheath tumors.

BACKGROUND: Malignant peripheral nerve sheath tumors (MPNST) are rare highly aggressive sarcomas that affect 8-13% of people with neurofibromatosis type 1. The prognosis for patients with MPNST is very poor. Despite TOP2A overexpression in these tumors, doxorubicin resistance is common, and the mechanisms of chemotherapy resistance in MPNST are poorly understood. Molecular-guided therapy prediction is an emerging strategy for treatment refractory sarcomas that involves identification of therapy response and resistance mechanisms in individual tumors. Here, we report the results from a personalized, molecular-guided therapy analysis of MPNST samples. METHODS: Established molecular-guided therapy prediction software algorithms were used to analyze published microarray data from human MPNST samples and cell lines, with benign neurofibroma tissue controls. MPNST and benign neurofibroma-derived cell lines were used for confirmatory in vitro experimentation using quantitative real-time PCR and growth inhibition assays. Microarray data was analyzed using Affymetrix expression console MAS 5.0 method. Significance was calculated with Welch's t-test with non-corrected p-value < 0.05 and validated using permutation testing across samples. Paired Student's t-tests were used to compare relative EC50 values from independent growth inhibition experiments. RESULTS: Molecular guided therapy predictions highlight substantial variability amongst human MPNST samples in expression of drug target and drug resistance pathways, as well as some similarities amongst samples, including common up-regulation of DNA repair mechanisms. In a subset of MPNSTs, high expression of ABCC1 is observed, serving as a predicted contra-indication for doxorubicin and related therapeutics in these patients. These microarray-based results are confirmed with quantitative, real-time PCR and immunofluorescence. The functional effect of drug efflux in MPNST-derived cells is confirmed using in vitro growth inhibition assays. Alternative therapeutics supported by the molecular-guided therapy predictions are reported and tested in MPNST-derived cells. CONCLUSIONS: These results confirm the substantial molecular heterogeneity of MPNSTs and validate molecular-guided therapy predictions in vitro. The observed molecular heterogeneity in MPNSTs influences therapy prediction. Also, mechanisms involving drug transport and DNA damage repair are primary mediators of MPNST chemotherapy resistance. Together, these findings support the utility of individualized therapy in MPNST as in other sarcomas, and provide initial proof-of concept that individualized therapy prediction can be accomplished.

A multi-site feasibility study for personalized medicine in canines with osteosarcoma.

BACKGROUND: A successful therapeutic strategy, specifically tailored to the molecular constitution of an individual and their disease, is an ambitious objective of modern medicine. In this report, we highlight a feasibility study in canine osteosarcoma focused on refining the infrastructure and processes required for prospective clinical trials using a series of gene expression-based Personalized Medicine (PMed) algorithms to predict suitable therapies within 5 days of sample receipt. METHODS: Tumor tissue samples were collected immediately following limb amputation and shipped overnight from veterinary practices. Upon receipt (day 1), RNA was extracted from snap-frozen tissue, with an adjacent H&E section for pathological diagnosis. Samples passing RNA and pathology QC were shipped to a CLIA-certified laboratory for genomic profiling. After mapping of canine probe sets to human genes and normalization against a (normal) reference set, gene level Z-scores were submitted to the PMed algorithms. The resulting PMed report was immediately forwarded to the veterinarians. Upon receipt and review of the PMed report, feedback from the practicing veterinarians was captured. RESULTS: 20 subjects were enrolled over a 5 month period. Tissue from 13 subjects passed both histological and RNA QC and were submitted for genomic analysis and subsequent PMed analysis and report generation. 11 of the 13 samples for which PMed reports were produced were communicated to the veterinarian within the target 5 business days. Of the 7 samples that failed QC, 4 were due to poor RNA quality, whereas 2 were failed following pathological review. Comments from the practicing veterinarians were generally positive and constructive, highlighting a number of areas for improvement, including enhanced education regarding PMed report interpretation, drug availability, affordable pricing and suitable canine dosing. CONCLUSIONS: This feasibility trial demonstrated that with the appropriate infrastructure and processes it is possible to perform an in-depth molecular analysis of a patient's tumor in support of real time therapeutic decision making within 5 days of sample receipt. A number of areas for improvement have been identified that should reduce the level of sample attrition and support clinical decision making.

Genomic characterization of explant tumorgraft models derived from fresh patient tumor tissue.

BACKGROUND: There is resurgence within drug and biomarker development communities for the use of primary tumorgraft models as improved predictors of patient tumor response to novel therapeutic strategies. Despite perceived advantages over cell line derived xenograft models, there is limited data comparing the genotype and phenotype of tumorgrafts to the donor patient tumor, limiting the determination of molecular relevance of the tumorgraft model. This report directly compares the genomic characteristics of patient tumors and the derived tumorgraft models, including gene expression, and oncogenic mutation status. METHODS: Fresh tumor tissues from 182 cancer patients were implanted subcutaneously into immune-compromised mice for the development of primary patient tumorgraft models. Histological assessment was performed on both patient tumors and the resulting tumorgraft models. Somatic mutations in key oncogenes and gene expression levels of resulting tumorgrafts were compared to the matched patient tumors using the OncoCarta (Sequenom, San Diego, CA) and human gene microarray (Affymetrix, Santa Clara, CA) platforms respectively. The genomic stability of the established tumorgrafts was assessed across serial in vivo generations in a representative subset of models. The genomes of patient tumors that formed tumorgrafts were compared to those that did not to identify the possible molecular basis to successful engraftment or rejection. RESULTS: Fresh tumor tissues from 182 cancer patients were implanted into immune-compromised mice with forty-nine tumorgraft models that have been successfully established, exhibiting strong histological and genomic fidelity to the originating patient tumors. Comparison of the transcriptomes and oncogenic mutations between the tumorgrafts and the matched patient tumors were found to be stable across four tumorgraft generations. Not only did the various tumors retain the differentiation pattern, but supporting stromal elements were preserved. Those genes down-regulated specifically in tumorgrafts were enriched in biological pathways involved in host immune response, consistent with the immune deficiency status of the host. Patient tumors that successfully formed tumorgrafts were enriched for cell signaling, cell cycle, and cytoskeleton pathways and exhibited evidence of reduced immunogenicity. CONCLUSIONS: The preservation of the patient's tumor genomic profile and tumor microenvironment supports the view that primary patient tumorgrafts provide a relevant model to support the translation of new therapeutic strategies and personalized medicine approaches in oncology.

Discordant inheritance of chromosomal and extrachromosomal DNA elements contributes to dynamic disease evolution in glioblastoma.

To understand how genomic heterogeneity of glioblastoma (GBM) contributes to poor therapy response, we performed DNA and RNA sequencing on GBM samples and the neurospheres and orthotopic xenograft models derived from them. We used the resulting dataset to show that somatic driver alterations including single-nucleotide variants, focal DNA alterations and oncogene amplification on extrachromosomal DNA (ecDNA) elements were in majority propagated from tumor to model systems. In several instances, ecDNAs and chromosomal alterations demonstrated divergent inheritance patterns and clonal selection dynamics during cell culture and xenografting. We infer that ecDNA was unevenly inherited by offspring cells, a characteristic that affects the oncogenic potential of cells with more or fewer ecDNAs. Longitudinal patient tumor profiling found that oncogenic ecDNAs are frequently retained throughout the course of disease. Our analysis shows that extrachromosomal elements allow rapid increase of genomic heterogeneity during GBM evolution, independently of chromosomal DNA alterations.

Cabozantinib (XL184) Inhibits Growth and Invasion of Preclinical TNBC Models.

PURPOSE: Triple-negative breast cancer (TNBC) is an aggressive breast cancer subtype that is associated with poor clinical outcome. There is a vital need for effective targeted therapeutics for TNBC patients, yet treatment strategies are challenged by the significant intertumoral heterogeneity within the TNBC subtype and its surrounding microenvironment. Receptor tyrosine kinases (RTK) are highly expressed in several TNBC subtypes and are promising therapeutic targets. In this study, we targeted the MET receptor, which is highly expressed across several TNBC subtypes. EXPERIMENTAL DESIGN: Using the small-molecule inhibitor cabozantinib (XL184), we examined the efficacy of MET inhibition in preclinical models that recapitulate human TNBC and its microenvironment. To analyze the dynamic interactions between TNBC cells and fibroblasts over time, we utilized a 3D model referred to as MAME (Mammary Architecture and Microenvironment Engineering) with quantitative image analysis. To investigate cabozantinib inhibition in vivo, we used a novel xenograft model that expresses human HGF and supports paracrine MET signaling. RESULTS: XL184 treatment of MAME cultures of MDA-MB-231 and HCC70 cells (± HGF-expressing fibroblasts) was cytotoxic and significantly reduced multicellular invasive outgrowths, even in cultures with HGF-expressing fibroblasts. Treatment with XL184 had no significant effects on MET(neg) breast cancer cell growth. In vivo assays demonstrated that cabozantinib treatment significantly inhibited TNBC growth and metastasis. CONCLUSIONS: Using preclinical TNBC models that recapitulate the breast tumor microenvironment, we demonstrate that cabozantinib inhibition is an effective therapeutic strategy in several TNBC subtypes.

Melanoma patient derived xenografts acquire distinct Vemurafenib resistance mechanisms.

Variable clinical responses, tumor heterogeneity, and drug resistance reduce long-term survival outcomes for metastatic melanoma patients. To guide and accelerate drug development, we characterized tumor responses for five melanoma patient derived xenograft models treated with Vemurafenib. Three BRAF(V600E) models showed acquired drug resistance, one BRAF(V600E) model had a complete and durable response, and a BRAF(V600V) model was expectedly unresponsive. In progressing tumors, a variety of resistance mechanisms to BRAF inhibition were uncovered, including mutant BRAF alternative splicing, NRAS mutation, COT (MAP3K8) overexpression, and increased mutant BRAF gene amplification and copy number. The resistance mechanisms among the patient derived xenograft models were similar to the resistance pathways identified in clinical specimens from patients progressing on BRAF inhibitor therapy. In addition, there was both inter- and intra-patient heterogeneity in resistance mechanisms, accompanied by heterogeneous pERK expression immunostaining profiles. MEK monotherapy of Vemurafenib-resistant tumors caused toxicity and acquired drug resistance. However, tumors were eradicated when Vemurafenib was combined the MEK inhibitor. The diversity of drug responses among the xenograft models; the distinct mechanisms of resistance; and the ability to overcome resistance by the addition of a MEK inhibitor provide a scheduling rationale for clinical trials of next-generation drug combinations.

Mesenchymal-like pancreatic cancer cells harbor specific genomic alterations more frequently than their epithelial-like counterparts.

The aggressiveness of pancreatic cancer is associated with the acquisition of mesenchymal characteristics by a subset of pancreatic cancer cells. The factors driving the development of this subset are not well understood. In this study, we tested the hypothesis that acquisition of a mesenchymal phenotype occurs selectively in tumor cells that harbor specific enabling genetic alterations. We obtained whole-genome comparative genomic hybridization (CGH) measurements on pancreatic cancer cell lines that have either an epithelial-like (17 cell lines) or a mesenchymal-like (9 cell lines) phenotype in vitro. The total amounts of amplifications and deletions were equivalent between the epithelial and mesenchymal groups, but 20 genes showed a major difference between the groups in prevalence of alterations. All 20 alterations (18 deletions and 2 amplifications) were more prevalent in the mesenchymal group, confirming the advanced nature of this cellular subtype. CDKN2A was altered in more than 50% of both groups, but co-deletions in neighboring genes, and concomitant loss of gene expression, were more prevalent in the mesenchymal group, suggesting that the size of the loss around CDKN2A affects cell phenotype. Whole-genome CGH on 11 primary cancer tissues revealed that the 20 genes were altered at a higher prevalence (up to 55% of the cases for certain genes) than randomly selected sets of 20 genes, with the same direction of alteration as in the cell lines. These findings support the concept that specific genetic alterations enable phenotype plasticity and provide promising candidate genes for further research.

Sox2 promotes malignancy in glioblastoma by regulating plasticity and astrocytic differentiation.

The high-mobility group-box transcription factor sex-determining region Y-box 2 (Sox2) is essential for the maintenance of stem cells from early development to adult tissues. Sox2 can reprogram differentiated cells into pluripotent cells in concert with other factors and is overexpressed in various cancers. In glioblastoma (GBM), Sox2 is a marker of cancer stemlike cells (CSCs) in neurosphere cultures and is associated with the proneural molecular subtype. Here, we report that Sox2 expression pattern in GBM tumors and patient-derived mouse xenografts is not restricted to a small percentage of cells and is coexpressed with various lineage markers, suggesting that its expression extends beyond CSCs to encompass more differentiated neoplastic cells across molecular subtypes. Employing a CSC derived from a patient with GBM and isogenic differentiated cell model, we show that Sox2 knockdown in the differentiated state abolished dedifferentiation and acquisition of CSC phenotype. Furthermore, Sox2 deficiency specifically impaired the astrocytic component of a biphasic gliosarcoma xenograft model while allowing the formation of tumors with sarcomatous phenotype. The expression of genes associated with stem cells and malignancy were commonly downregulated in both CSCs and serum-differentiated cells on Sox2 knockdown. Genes previously shown to be associated with pluripontency and CSCs were only affected in the CSC state, whereas embryonic stem cell self-renewal genes and cytokine signaling were downregulated, and the Wnt pathway activated in differentiated Sox2-deficient cells. Our results indicate that Sox2 regulates the expression of key genes and pathways involved in GBM malignancy, in both cancer stemlike and differentiated cells, and maintains plasticity for bidirectional conversion between the two states, with significant clinical implications.

Prospective molecular profiling of canine cancers provides a clinically relevant comparative model for evaluating personalized medicine (PMed) trials.

BACKGROUND: Molecularly-guided trials (i.e. PMed) now seek to aid clinical decision-making by matching cancer targets with therapeutic options. Progress has been hampered by the lack of cancer models that account for individual-to-individual heterogeneity within and across cancer types. Naturally occurring cancers in pet animals are heterogeneous and thus provide an opportunity to answer questions about these PMed strategies and optimize translation to human patients. In order to realize this opportunity, it is now necessary to demonstrate the feasibility of conducting molecularly-guided analysis of tumors from dogs with naturally occurring cancer in a clinically relevant setting. METHODOLOGY: A proof-of-concept study was conducted by the Comparative Oncology Trials Consortium (COTC) to determine if tumor collection, prospective molecular profiling, and PMed report generation within 1 week was feasible in dogs. Thirty-one dogs with cancers of varying histologies were enrolled. Twenty-four of 31 samples (77%) successfully met all predefined QA/QC criteria and were analyzed via Affymetrix gene expression profiling. A subsequent bioinformatics workflow transformed genomic data into a personalized drug report. Average turnaround from biopsy to report generation was 116 hours (4.8 days). Unsupervised clustering of canine tumor expression data clustered by cancer type, but supervised clustering of tumors based on the personalized drug report clustered by drug class rather than cancer type. CONCLUSIONS: Collection and turnaround of high quality canine tumor samples, centralized pathology, analyte generation, array hybridization, and bioinformatic analyses matching gene expression to therapeutic options is achievable in a practical clinical window (<1 week). Clustering data show robust signatures by cancer type but also showed patient-to-patient heterogeneity in drug predictions. This lends further support to the inclusion of a heterogeneous population of dogs with cancer into the preclinical modeling of personalized medicine. Future comparative oncology studies optimizing the delivery of PMed strategies may aid cancer drug development.

Comparative RNA-Seq and microarray analysis of gene expression changes in B-cell lymphomas of Canis familiaris.

Comparative oncology is a developing research discipline that is being used to assist our understanding of human neoplastic diseases. Companion canines are a preferred animal oncology model due to spontaneous tumor development and similarity to human disease at the pathophysiological level. We use a paired RNA sequencing (RNA-Seq)/microarray analysis of a set of four normal canine lymph nodes and ten canine lymphoma fine needle aspirates to identify technical biases and variation between the technologies and convergence on biological disease pathways. Surrogate Variable Analysis (SVA) provides a formal multivariate analysis of the combined RNA-Seq/microarray data set. Applying SVA to the data allows us to decompose variation into contributions associated with transcript abundance, differences between the technology, and latent variation within each technology. A substantial and highly statistically significant component of the variation reflects transcript abundance, and RNA-Seq appeared more sensitive for detection of transcripts expressed at low levels. Latent random variation among RNA-Seq samples is also distinct in character from that impacting microarray samples. In particular, we observed variation between RNA-Seq samples that reflects transcript GC content. Platform-independent variable decomposition without a priori knowledge of the sources of variation using SVA represents a generalizable method for accomplishing cross-platform data analysis. We identified genes differentially expressed between normal lymph nodes of disease free dogs and a subset of the diseased dogs diagnosed with B-cell lymphoma using each technology. There is statistically significant overlap between the RNA-Seq and microarray sets of differentially expressed genes. Analysis of overlapping genes in the context of biological systems suggests elevated expression and activity of PI3K signaling in B-cell lymphoma biopsies compared with normal biopsies, consistent with literature describing successful use of drugs targeting this pathway in lymphomas.

MEK2 is sufficient but not necessary for proliferation and anchorage-independent growth of SK-MEL-28 melanoma cells.

Mitogen-activated protein kinase kinases (MKK or MEK) 1 and 2 are usually treated as redundant kinases. However, in assessing their relative contribution towards ERK-mediated biologic response investigators have relied on tests of necessity, not sufficiency. In response we developed a novel experimental model using lethal toxin (LeTx), an anthrax toxin-derived pan-MKK protease, and genetically engineered protease resistant MKK mutants (MKKcr) to test the sufficiency of MEK signaling in melanoma SK-MEL-28 cells. Surprisingly, ERK activity persisted in LeTx-treated cells expressing MEK2cr but not MEK1cr. Microarray analysis revealed non-overlapping downstream transcriptional targets of MEK1 and MEK2, and indicated a substantial rescue effect of MEK2cr on proliferation pathways. Furthermore, LeTx efficiently inhibited the cell proliferation and anchorage-independent growth of SK-MEL-28 cells expressing MKK1cr but not MEK2cr. These results indicate in SK-MEL-28 cells MEK1 and MEK2 signaling pathways are not redundant and interchangeable for cell proliferation. We conclude that in the absence of other MKK, MEK2 is sufficient for SK-MEL-28 cell proliferation. MEK1 conditionally compensates for loss of MEK2 only in the presence of other MKK.

MicroRNA 92a-2*: a biomarker predictive for chemoresistance and prognostic for survival in patients with small cell lung cancer.

PURPOSE: Although the majority of patients with small cell lung cancer (SCLC) respond to initial chemotherapy, those with disease progression at first response assessment (chemoresistance) have inferior outcomes. There is a need for predictive biomarkers to aid investigators in designing future clinical trials that better stratify patients beyond standard clinical and laboratory parameters and to identify new treatments for this patient subpopulation. We hypothesized that tumor microRNAs (miRNAs) could serve as predictive biomarkers for chemoresistance and prognostic biomarkers for survival of patients with SCLC treated with systemic chemotherapy. PATIENTS AND METHODS: SCLC samples annotated with clinical characteristics and baseline comorbidities were available. miRNA microarray profiling was performed on diagnostic SCLC tumor samples, and analysis was performed using XenoBase, a data integration and discovery tool. Confirmation of the top 16 miRNA candidates was performed using quantitative real-time polymerase chain reaction followed by analyses to determine clinical and miRNA biomarkers associated with chemoresistance and survival. RESULTS: miRNAs significantly associated with chemoresistance were miR-92a-2* (p = 0.010), miR-147 (p = 0.018), and miR-574-5p (p = 0.039). By stepwise multivariate analysis, only gender and miR-92a-2* contributed significantly to survival (p = 0.023) and (p = 0.015), respectively. Baseline comorbidities were not associated with chemoresistance or survival. CONCLUSIONS: Higher tumor miR-92a-2* levels are associated with chemoresistance and with decreased survival in patients with SCLC. Tumor miR-92a-2* may have application in screening patients with SCLC at risk for de novo chemoresistance in an effort to design more tailored clinical trials for this subpopulation. Further validation in independent sample sets is warranted.

Identifying disease-specific genes based on their topological significance in protein networks.

BACKGROUND: The identification of key target nodes within complex molecular networks remains a common objective in scientific research. The results of pathway analyses are usually sets of fairly complex networks or functional processes that are deemed relevant to the condition represented by the molecular profile. To be useful in a research or clinical laboratory, the results need to be translated to the level of testable hypotheses about individual genes and proteins within the condition of interest. RESULTS: In this paper we describe novel computational methodology capable of predicting key regulatory genes and proteins in disease- and condition-specific biological networks. The algorithm builds shortest path network connecting condition-specific genes (e.g. differentially expressed genes) using global database of protein interactions from MetaCore. We evaluate the number of all paths traversing each node in the shortest path network in relation to the total number of paths going via the same node in the global network. Using these numbers and the relative size of the initial data set, we determine the statistical significance of the network connectivity provided through each node. We applied this method to gene expression data from psoriasis patients and identified many confirmed biological targets of psoriasis and suggested several new targets. Using predicted regulatory nodes we were able to reconstruct disease pathways that are in excellent agreement with the current knowledge on the pathogenesis of psoriasis. CONCLUSION: The systematic and automated approach described in this paper is readily applicable to uncovering high-quality therapeutic targets, and holds great promise for developing network-based combinational treatment strategies for a wide range of diseases.