Search

Eribulin, a natural product-based microtubule targeting agent with cytotoxic and noncytotoxic mechanisms, is FDA approved for certain patients with advanced breast cancer and liposarcoma. To investigate the feasibility of developing drug-specific predictive biomarkers, we quantified antiproliferative activities of eribulin versus paclitaxel and vinorelbine against 100 human cancer cell lines from the Cancer Cell Line Encyclopedia, and correlated results with publicly available databases to identify genes and pathways associated with eribulin response, either uniquely or shared with paclitaxel or vinorelbine. Mean expression ratios of 11,985 genes between the most and least sensitive cell line quartiles were sorted by p-values and drug overlaps, yielding 52, 29 and 80 genes uniquely associated with eribulin, paclitaxel and vinorelbine, respectively. Further restriction to minimum 2-fold ratios followed by reintroducing data from the middle two quartiles identified 9 and 13 drug-specific unique fingerprint genes for eribulin and vinorelbine, respectively; surprisingly, no gene met all criteria for paclitaxel. Interactome and Reactome pathway analyses showed that unique fingerprint genes of both drugs were primarily associated with cellular signaling, not microtubule-related pathways, although considerable differences existed in individual pathways identified. Finally, four-gene (C5ORF38, DAAM1, IRX2, CD70) and five-gene (EPHA2, NGEF, SEPTIN10, TRIP10, VSIG10) multivariate regression models for eribulin and vinorelbine showed high statistical correlation with drug-specific responses across the 100 cell lines and accurately calculated predicted mean IC50s for the most and least sensitive cell line quartiles as surrogates for responders and nonresponders, respectively. Collectively, these results provide a foundation for developing drug-specific predictive biomarkers for eribulin and vinorelbine.

Introduction: Immune Checkpoint Inhibition (ICI) therapies have demonstrated significant clinical success in multiple cancer types. However, many patients do not develop durable responses to ICI treatments and the underlying mechanisms of action are not fully resolved. Recent single cell RNA-seq (scRNA-seq) studies provide a window into the complex interplay between cancer and heterogenous immune cell populations underlying ICI response. However, individual studies are typically focused on specific therapies in specific cancer types making it challenging to identify universal mechanisms and response markers. Approach: To facilitate a more comprehensive study of cell-type specific ICI mechanisms and response determinants, we developed the Single Cell Immune Checkpoint Inhibition Atlas (ICI Atlas) which integrates 27 scRNA-seq datasets (including 17 human and 10 mice datasets) spanning close to 1.9 million cells. Of the 27 datasets, 25 contain both pre- and post-treatment samples, one contains only pre-treatment samples and one contains only post-treatment samples. The Atlas spans over 30 cell types and 16 different cancer types as well as 20 treatment strategies spanning 11 ICI therapies and 28 other drugs. Results: To begin exploring the cell-type specific responses to ICI, we conducted comparative analyses of the ICI Atlas T cells in 4 groups: 1) responders vs. non-responders pre-treatment; 2) responders vs. non-responders post-treatment; 3) responders post-treatment vs. untreated; and 4) non-responders post-treatment vs. untreated leading to 62 pairwise comparisons across ICI Atlas datasets and treatment strategies. First, we found that T cells were either more abundant or more differentiated after ICI treatment and were more differentiated or more abundant in responders (post-treatment) compared to non-responders. We also found that genes previously identified in a meta-analysis of bulk mRNA datasets (Litchfield et al. 2021) as being effective predictors of ICI response are also more highly expressed in T cells of responders post-treatment compared to non-responders or compared to untreated subjects. Using an IFN-related gene signature, we found that post treatment as compared to baseline, IFN genes increased more substantially in responders than in non-responders. Across a large set of studies, we found that the IFN signature clusters T cells into IFN-effector/producers (effector/activated T cells) and IFN-insensitive cells (naive/memory T cells). Finally, using Atlas-derived ICI response signatures in T cells and data from a recently published study of mismatch repair deficient colorectal cancers (MMRd CRC, Pelka et al. 2021), we showed that T cells from MMRd CRC patients recapitulate signatures of ICI responders. Our results suggest that large-scale scRNA-seq atlases can serve as valuable tools for corroborating existing hypotheses, as well as providing key new insights into cell-type specific mechanisms underlying ICI therapy response. Citation Format: Piotr Topolewski, Aleksandra Możwiłło, Marlena Osipowicz, Roy Ronen, Janusz Dutkowski. Interrogating the single cell immune checkpoint inhibition atlas to decipher ICI response mechanisms [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 2 (Clinical Trials and Late-Breaking Research); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(8_Suppl):Abstract nr LB176.

PURPOSE Patients with metastatic triple-negative breast cancer (mTNBC) have poor outcomes. The Intensive Trial of Omics in Cancer (ITOMIC) sought to determine the feasibility and potential efficacy of informing treatment decisions through multiple biopsies of mTNBC deposits longitudinally over time, accompanied by analysis using a distributed network of experts. METHODS Thirty-one subjects were enrolled and 432 postenrollment biopsies performed (clinical and study-directed) of which 332 were study-directed. Molecular profiling included whole-genome sequencing or whole-exome sequencing, cancer-associated gene panel sequencing, RNA-sequencing, and immunohistochemistry. To afford time for analysis, subjects were initially treated with cisplatin (19 subjects), or another treatment they had not received previously. The results were discussed at a multi-institutional ITOMIC Tumor Board, and a report transmitted to the subject's oncologist who arrived at the final treatment decision in conjunction with the subject. Assistance was provided to access treatments that were predicted to be effective. RESULTS Multiple biopsies in single settings and over time were safe, and comprehensive analysis was feasible. Two subjects were found to have lung cancer, one had carcinoma of unknown primary site, tumor samples from three subjects were estrogen receptor–positive and from two others, human epidermal growth factor receptor 2–positive. Two subjects withdrew. Thirty-four of 112 recommended treatments were accessed using approved drugs, clinical trials, and single-patient investigational new drugs. After excluding the three subjects with nonbreast cancers and the two subjects who withdrew, 22 of 26 subjects (84.6%) received at least one ITOMIC Tumor Board–recommended treatment. CONCLUSION Further exploration of this approach in patients with mTNBC is merited.

Background: Effective target discovery strategies are key to developing new precision medicines against cancer. Large-scale efforts to systematically characterize cancer-associated germline variants, genomic changes in tumors and cancer genetic dependencies via CRISPR and RNAi in cell lines provide valuable resources for machine learning (ML) approaches designed to identify disease-modifying targets. However, fundamental challenges remain in understanding how various data types (including primary tumor and cell line data) can be jointly modeled to decipher the molecular network contexts underpinning the biology and efficacy of new target candidates. Approach: Using a data-driven knowledge graph spanning >2 billion relations over >350,000 biomedical entities, we assembled a comprehensive target prioritization resource with 53 feature types including multi-omics (germline and tumor) and genetic dependency data. Feature categories consisted of raw feature types (e.g. germline genetic, somatic mutation, copy number variation, transcriptomic, proteomic, survival associations, CRISPR/RNAi sensitivity, selectivity, efficacy and essentiality), network-transformed features capturing the underlying molecular network context of each possible target, and multidimensional network integration features combining systems-level information from multiple feature types. Multiple ML models were trained on either omics-based features only or on both omics and genetic dependency features; and using information from each cancer type separately, as well as jointly in multi-disease models. Results: Individual features and the resulting ML models were evaluated using known drug targets in 20 cancer types. We found that a) germline genetic and survival association features along with genetic dependency (efficacy and selectivity) features were the most predictive prior to network transformation; b) network-based feature transformation raised the predictive power of individual features (e.g. increasing mean AUPRC from 0.024 to 0.064 for germline genetic and from 0.018 to 0.048 for cell line sensitivity features); c) network integration of multiple feature types increased mean AUPRC by 21% over the best individual omics features; d) ML models integrating both omics and genetic dependency data outperformed ML omics-only approaches (mean AUPRC 0.219 vs. 0.195); and e) multi-disease ML models provided the best overall performance (mean AUPRC 0.232) underlining the value of multi-dimensional data integration and information transfer across cancer types. Resulting network-based machine learning models provide a highly interpretable view into the top-scoring genes for each cancer, prioritizing established targets, targets recently being evaluated (e.g. FGFR2 and FGFR3 in lung squamous cell carcinoma) and novel target candidates. Citation Format: Janusz Dutkowski, Radosław Bielecki, Karol Nienałtowski, Michał Kukiełka, Roy Ronen. Knowledge graph integration of germline, primary cancer and cancer genetic dependency data prioritizes new target candidates [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr LB148.

Summary Accurately translating genotype to phenotype requires accounting for the functional impact of genetic variation at many biological scales. Here we present a strategy for genotype-phenotype reasoning based on existing knowledge of cellular subsystems. These subsystems and their hierarchical organization are defined by the Gene Ontology or a complementary ontology inferred directly from previously published datasets. Guided by the ontology’s hierarchical structure, we organize genotype data into an “ontotype,” that is, a hierarchy of perturbations representing the effects of genetic variation at multiple cellular scales. The ontotype is then interpreted using logical rules generated by machine learning to predict phenotype. This approach substantially outperforms previous, non-hierarchical methods for translating yeast genotype to cell growth phenotype, and it accurately predicts the growth outcomes of two new screens of 2,503 double gene knockouts impacting DNA repair or nuclear lumen. Ontotypes also generalize to larger knockout combinations, setting the stage for interpreting the complex genetics of disease.

Background: Eribulin mesylate, a synthetic analog of the marine sponge natural product halichondrin B, is a microtubule dynamics inhibitor that acts via both cytotoxic antimitotic mechanisms as well as non-cytotoxic effects on tumor vasculature, tumor phenotype and the tumor microenvironment. Eribulin is approved in the US for treatment of certain patients with advanced breast cancer or liposarcoma. Here we examined in vitro antiproliferative responses to eribulin versus vinorelbine and paclitaxel in 100 human cancer cell lines from the Cancer Cell Line Encyclopedia (CCLE), followed by analysis of CCLE baseline gene expression data with the goal of assessing overlapping and non-overlapping pathways and genes associated with drug response and resistance. Methods: In vitro proliferation assays were conducted on 100 CCLE cell lines. Cancer types represented by ≥5 cell lines include breast, lung, blood, colorectum, ovary, pancreas, skin and kidney. Comparative systems-level analyses of genes, networks and pathways associated with drug response and/or resistance were performed to identify gene and pathway-level determinants of drug response that were either unique to eribulin or shared with vinorelbine and/or paclitaxel. Results: We evaluated genes and molecular pathways driving sensitivity to eribulin versus vinorelbine and paclitaxel in a panel of 100 CCLE cell lines and identified key molecular determinants of response and resistance unique to eribulin. Pathway expression analysis revealed key pathways significantly associated with response to eribulin, including those associated with oxidative stress-induced senescence, histone deacetylation, DNA methylation and Notch signaling. The oxidative stress-induced senescence pathway was significantly enriched for genes whose pretreatment/basal mRNA expression levels were positively associated with response to eribulin but not vinorelbine or paclitaxel. Expression levels of various tubulin isoforms were also identified as positively correlated with eribulin response. Higher and lower pretreatment expression of E-cadherin and N-cadherin, respectively, were also found in cell lines sensitive to eribulin treatment. Conclusion: Pretreatment gene expression analysis demonstrated key gene sets and molecular pathways that may drive response and resistance to eribulin and may differentiate this drug from vinorelbine and paclitaxel. Detailed analysis will be presented at the meeting. Citation Format: Pallavi Sachdev, Roy Ronen, Janusz Dutkowski, Bruce A. Littlefield. Genetic analysis of responses to eribulin versus vinorelbine and paclitaxel in 100 cancer cell lines from the cancer cell line Encyclopedia (CCLE) [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 1924.

Background As Kawasaki disease (KD) shares many clinical features with other more common febrile illnesses and misdiagnosis, leading to a delay in treatment, increases the risk of coronary artery damage, a diagnostic test for KD is urgently needed. We sought to develop a panel of biomarkers that could distinguish between acute KD patients and febrile controls (FC) with sufficient accuracy to be clinically useful. Methods Plasma samples were collected from three independent cohorts of FC and acute KD patients who met the American Heart Association definition for KD and presented within the first 10 days of fever. The levels of 88 biomarkers associated with inflammation were assessed by Luminex bead technology. Unsupervised clustering followed by supervised clustering using a Random Forest model was used to find a panel of candidate biomarkers. Results A panel of biomarkers commonly available in the hospital laboratory (absolute neutrophil count, erythrocyte sedimentation rate, alanine aminotransferase, gamma glutamyl transferase, concentrations of alpha-1-antitrypsin, C-reactive protein, and fibrinogen, and platelet count) accurately diagnosed 81 to 96% of KD patients in a series of three independent cohorts. Conclusions After prospective validation, this 8-biomarker panel may improve the recognition of KD.

Motivation: While the manually curated Gene Ontology (GO) is widely used, inferring a GO directly from -omics data is a compelling new problem. Recognizing that ontologies are a directed acyclic graph (DAG) of terms and hierarchical relations, algorithms are needed that: analyze a full matrix of gene–gene pairwise similarities from -omics data;infer true hierarchical structure in these data rather than enforcing hierarchy as a computational artifact; andrespect biological pleiotropy, by which a term in the hierarchy can relate to multiple higher level terms. Methods addressing these requirements are just beginning to emerge—none has been evaluated for GO inference. Methods: We consider two algorithms [Clique Extracted Ontology (CliXO), LocalFitness] that uniquely satisfy these requirements, compared with methods including standard clustering. CliXO is a new approach that finds maximal cliques in a network induced by progressive thresholding of a similarity matrix. We evaluate each method’s ability to reconstruct the GO biological process ontology from a similarity matrix based on (a) semantic similarities for GO itself or (b) three -omics datasets for yeast. Results: For task (a) using semantic similarity, CliXO accurately reconstructs GO (>99% precision, recall) and outperforms other approaches ( Conclusion: This study provides algorithmic foundation for building gene ontologies by capturing hierarchical and pleiotropic structure embedded in biomolecular data. Contact: tideker@ucsd.edu

Background: Detailed understanding of cancer subtypes is key to precision oncology and instrumental for effective patient-to-therapy assignment and development of combination therapy regimens. While many molecular and pathologic cancer subtypes have been developed over the last decade, key challenges remain. Paramount among them is the ability to effectively track and compare cancer subtypes across studies, identify meaningful associations with clinical outcomes, and robustly assign new patients into subtypes in a clinical setting. Approach: Here we report on the development of the Cancer Subtype Ontology (CSO) - a global resource for cancer subtype interrogation and patient-to-subtype assignment, constituting, to our knowledge, the largest collection of cancer subtypes to date. The foundational CSO resource currently catalogues 840 cancer subtypes which subcategorize ~40 histologies based on a diverse range of evidence types ranging from cancer genomics and transcriptomics, to epigenetics, proteomics and immune infiltration. CSO includes a wide collection of previously-published subtypes as well as a new model-based subtype collection reported here for the first time. CSO provides tools to compare across subtypes as well as characterize subtype enrichment for molecular, pathologic and clinical features, including response to therapy and patient survival. As an integrative part of the CSO, we develop a novel machine learning framework for robust assignment of new clinical and pre-clinical samples into the CSO subtypes. We first evaluate the CSO subtype collection using TCGA, measuring the ability to predict overall, progression-free, disease-specific, and disease-free survival. Next, using a novel curated collection of recent immune checkpoint inhibitor (ICI) trials, we demonstrate the power of the subtypes to predict ICI treatment responses and clinical outcomes. Results: First using TCGA, we found that the CSO significantly expands the ability to uncover subtypes predictive of clinical outcomes. Notably, new model-based subtypes in CSO, while constituting less than 1/3 of the ontology, accounted for ~60% of the most significant outcome associations across genomic and molecular subtypings. Next, we applied the CSO patient assignment system to systematically assign patients from 10 ICI trials across 4 histologies into 103 CSO subtypes. We found that half of the trials had CSO assignments predictive of patient survival and/or response and most of them were novel. Specifically, assignment of IMvigor 210 urinary bladder cancer patients revealed two subtypes enriched for partial (P=0.0002, OR=3.87) and complete response (P =0.002, OR=6.86), respectively, only one of which was previously characterized (Mariathasan et al. 2018). For the Snyder et al. 2014 cohort we observed that while stratification using the cohort data alone did not yield predictive subtypes, CSO patient assignment revealed subtypes that were predictive of long term clinical benefit (P=0.003). Furthermore, across non-small cell lung cancer trials analyzed, CSO assignment was predictive of progression-free survival (P=0.048) in the Rizvi et al. 2018 cohort and showed suggestive associations with both progression-free survival (P=0.055) and response (durable clinical benefit, P=0.08) for the Rizvi et al. 2016 cohort. Citation Format: Roy Ronen, Christopher DeBoever, Boguslaw Kluge, Janusz Dutkowski. The cancer subtype ontology: Predicting response to therapy in clinical and translational research using 840 cancer subtypes [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr LB-217.

Comparative approaches in genomics have long relied on rigorous mathematical models of sequence evolution. Such models provide the basis for formulating and solving well-defined computational problems, in turn yielding key insights into the evolutionary processes acting on the genome. Analogous model-based approaches for analyzing biological networks are still under development. Here we describe a model-based approach for estimating the probability of network rewiring events during evolution. Our method builds on the standard duplication-and-divergence model and incorporates phylogenetic analysis to guide the comparison of protein networks across species. We apply our algorithm to study the evolution of functional modules and unconstrained network regions in seven available eukaryotic interactomes. Based on this analysis we identify a map of co-functioning protein families whose members participate in strongly conserved interactions and form major complexes and pathways in the eukaryotic cell. The proposed approach provides principled means for inferring the probability of network rewiring events, enabling insights into the conservation and divergence of protein interactions and the formation of functional modules in protein networks.

Accelerating cancer research is expected to require new types of clinical trials. This report describes the Intensive Trial of OMics in Cancer (ITOMIC) and a participant with triple-negative breast cancer metastatic to bone, who had markedly elevated circulating tumor cells (CTCs) that were monitored 48 times over 9 months. A total of 32 researchers from 14 institutions were engaged in the patient's evaluation; 20 researchers had no prior involvement in patient care and 18 were recruited specifically for this patient. Whole-exome sequencing of 3 bone marrow samples demonstrated a novel ROS1 variant that was estimated to be present in most or all tumor cells. After an initial response to cisplatin, a hypothesis of crizotinib sensitivity was disproven. Leukapheresis followed by partial CTC enrichment allowed for the development of a differential high-throughput drug screen and demonstrated sensitivity to investigational BH3-mimetic inhibitors of BCL-2 that could not be tested in the patient because requests to the pharmaceutical sponsors were denied. The number and size of CTC clusters correlated with clinical status and eventually death. Focusing the expertise of a distributed network of investigators on an intensively monitored patient with cancer can generate high-resolution views of the natural history of cancer and suggest new opportunities for therapy. Optimization requires access to investigational drugs.

A major difficulty in the analysis of complex biological systems is dealing with the low signal-to-noise inherent to nearly all large biological datasets. We discuss powerful bioinformatic concepts for boosting signal-to-noise through external knowledge incorporated in processing units we call filters and integrators. These concepts are illustrated in four landmark studies that have provided model implementations of filters, integrators, or both.

Background: Immunotherapies are revolutionizing cancer care but identifying patients which are most likely to achieve clinical benefit remains a key challenge. Available predictors such as PD-1/PD-L1 expression, tumor mutation burden or MSI-H/d-MMR status have shown promise in predicting response in specific cancer types, they fail to explain or predict response for many patients. A more comprehensive understating of biology underlying response to therapy is needed. Approach: We developed an integrated machine-learning system that uses a mathematical model of cancer subtypes and a neoantigen fitness model to stratify multimodal omics data along with associated profiles of immune infiltration and immune pathway activity. The system holistically determines the underlying pathway-level architecture of genomic and other molecular alterations in each cancer. It further uses a new immune pathway deconvolution approach to determine subtype-specific immune-infiltrates and activation of key immune and immune-modulating pathways that may be associated with response to therapy. Results: The system is applied to stratify and interrogate subtypes in over 40 cancer types. Here we discuss results for two of the histologies for which immunotherapies are being pursued clinically. Comprehensive systems-level interrogation of tumor genomic architecture and tumor microenvironment leads to new molecular sub-typings for colorectal adenocarcinoma and skin cutaneous melanoma, here referred to as COADREAD_MUT_CNV_MODEL_6 and SKCM_MUT_CNV_MODEL_3, respectively. The model-based subtypes uncover critical patterns of co-occurring and mutually-exclusive somatic alterations not apparent using previous approaches, and show strong correlation with immune infiltration. Further integration of the subtyping model with a new neoantigen fitness model significantly improves performance in predicting response to checkpoint inhibition compared to existing methods. Citation Format: Roy Ronen, Boguslaw Kluge, Janusz Dutkowski. Deep genomic, immune-infiltrate and neoantigen-based stratification of cancer predicts response to immune checkpoint inhibition [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr LB-028.

It is now established that the transcription factors E2A, EBF1 and Foxo1 have critical roles in B cell development. Here we show that E2A and EBF1 bound regulatory elements present in the Foxo1 locus. E2A and EBF1, as well as E2A and Foxo1, in turn, were wired together by a vast spectrum of cis-regulatory sequences. These associations were dynamic during developmental progression. Occupancy by the E2A isoform E47 directly resulted in greater abundance, as well as a pattern of monomethylation of histone H3 at lysine 4 (H3K4) across putative enhancer regions. Finally, we divided the pro-B cell epigenome into clusters of loci with occupancy by E2A, EBF and Foxo1. From this analysis we constructed a global network consisting of transcriptional regulators, signaling and survival factors that we propose orchestrates B cell fate.

Here we develop a fully automated procedure for the analysis of liquid chromatography–mass spectrometry (LC–MS) datasets collected during the analysis of complex peptide mixtures. We present the underlying algorithm and outcomes of several experiments justifying its applicability. The novelty of our approach is to exploit the multidimensional character of the datasets. It is common knowledge that highly complex peptide mixtures can be analyzed by liquid chromatography coupled with mass spectrometry, but we are not aware of any existing automated MS spectra interpretation procedure designed to take into account the multidimensional character of the data. Our work fills this gap by providing an effective algorithm for this task, allowing for automated conversion of raw data to the list of masses of peptides.

Although it is well-established that the macrophage M1 to M2 transition plays a role in tumor progression, the molecular basis for this process remains incompletely understood. Herein, we demonstrate that the small GTPase, Rac2 controls macrophage M1 to M2 differentiation and the metastatic phenotype in vivo. Using a genetic approach, combined with syngeneic and orthotopic tumor models we demonstrate that Rac2-/- mice display a marked defect in tumor growth, angiogenesis and metastasis. Microarray, RT-PCR and metabolomic analysis on bone marrow derived macrophages isolated from the Rac2-/- mice identify an important role for Rac2 in M2 macrophage differentiation. Furthermore, we define a novel molecular mechanism by which signals transmitted from the extracellular matrix via the α4β1 integrin and MCSF receptor lead to the activation of Rac2 and potentially regulate macrophage M2 differentiation. Collectively, our findings demonstrate a macrophage autonomous process by which the Rac2 GTPase is activated downstream of the α4β1 integrin and the MCSF receptor to control tumor growth, metastasis and macrophage differentiation into the M2 phenotype. Finally, using gene expression and metabolomic data from our Rac2-/- model, and information related to M1-M2 macrophage differentiation curated from the literature we executed a systems biologic analysis of hierarchical protein-protein interaction networks in an effort to develop an iterative interactome map which will predict additional mechanisms by which Rac2 may coordinately control macrophage M1 to M2 differentiation and metastasis.

The Network-extracted Ontology (NeXO) is a gene ontology inferred directly from large-scale molecular networks. While most ontologies are constructed through manual expert curation, NeXO uses a principled computational approach which integrates evidence from hundreds of thousands of individual gene and protein interactions to construct a global hierarchy of cellular components and processes. Here, we describe the development of the NeXO Web platform (http://www.nexontology.org)-an online database and graphical user interface for visualizing, browsing and performing term enrichment analysis using NeXO and the gene ontology. The platform applies state-of-the-art web technology and visualization techniques to provide an intuitive framework for investigating biological machinery captured by both data-driven and manually curated ontologies.

Breast cancer is the most common malignancy in women and is responsible for hundreds of thousands of deaths annually. As with most cancers, it is a heterogeneous disease and different breast cancer subtypes are treated differently. Understanding the difference in prognosis for breast cancer based on its molecular and phenotypic features is one avenue for improving treatment by matching the proper treatment with molecular subtypes of the disease. In this work, we employed a competition-based approach to modeling breast cancer prognosis using large datasets containing genomic and clinical information and an online real-time leaderboard program used to speed feedback to the modeling team and to encourage each modeler to work towards achieving a higher ranked submission. We find that machine learning methods combined with molecular features selected based on expert prior knowledge can improve survival predictions compared to current best-in-class methodologies and that ensemble models trained across multiple user submissions systematically outperform individual models within the ensemble. We also find that model scores are highly consistent across multiple independent evaluations. This study serves as the pilot phase of a much larger competition open to the whole research community, with the goal of understanding general strategies for model optimization using clinical and molecular profiling data and providing an objective, transparent system for assessing prognostic models., Author Summary We developed an extensible software framework for sharing molecular prognostic models of breast cancer survival in a transparent collaborative environment and subjecting each model to automated evaluation using objective metrics. The computational framework presented in this study, our detailed post-hoc analysis of hundreds of modeling approaches, and the use of a novel cutting-edge data resource together represents one of the largest-scale systematic studies to date assessing the factors influencing accuracy of molecular-based prognostic models in breast cancer. Our results demonstrate the ability to infer prognostic models with accuracy on par or greater than previously reported studies, with significant performance improvements by using state-of-the-art machine learning approaches trained on clinical covariates. Our results also demonstrate the difficultly in incorporating molecular data to achieve substantial performance improvements over clinical covariates alone. However, improvement was achieved by combining clinical feature data with intelligent selection of important molecular features based on domain-specific prior knowledge. We observe that ensemble models aggregating the information across many diverse models achieve among the highest scores of all models and systematically out-perform individual models within the ensemble, suggesting a general strategy for leveraging the wisdom of crowds to develop robust predictive models.

Ontologies have proven very useful for capturing knowledge as a hierarchy of terms and their interrelationships. In biology a major challenge has been to construct ontologies of gene function given incomplete biological knowledge and inconsistencies in how this knowledge is manually curated. Here we show that large networks of gene and protein interactions in Saccharomyces cerevisiae can be used to infer an ontology whose coverage and power are equivalent to those of the manually curated Gene Ontology (GO). The network-extracted ontology (NeXO) contains 4,123 biological terms and 5,766 term-term relations, capturing 58% of known cellular components. We also explore robust NeXO terms and term relations that were initially not cataloged in GO, a number of which have now been added based on our analysis. Using quantitative genetic interaction profiling and chemogenomics, we find further support for many of the uncharacterized terms identified by NeXO, including multisubunit structures related to protein trafficking or mitochondrial function. This work enables a shift from using ontologies to evaluate data to using data to construct and evaluate ontologies.

SummaryRecent studies have documented genome-wide binding patterns of transcriptional regulators and their associated epigenetic marks in hematopoietic cell lineages. In order to determine how epigenetic marks are established and maintained during developmental progression, we have generated long-term cultures of hematopoietic progenitors by enforcing the expression of the E-protein antagonist Id2. Hematopoietic progenitors that express Id2 are multipotent and readily differentiate upon withdrawal of Id2 expression into committed B lineage cells, thus indicating a causative role for E2A (Tcf3) in promoting the B cell fate. Genome-wide analyses revealed that a substantial fraction of lymphoid and myeloid enhancers are premarked by the poised or active enhancer mark H3K4me1 in multipotent progenitors. Thus, in hematopoietic progenitors, multilineage priming of enhancer elements precedes commitment to the lymphoid or myeloid cell lineages.

Many biological and clinical outcomes are based not on single proteins, but on modules of proteins embedded in protein networks. A fundamental question is how the proteins within each module contribute to the overall module activity. Here, we study the modules underlying three representative biological programs related to tissue development, breast cancer metastasis, or progression of brain cancer, respectively. For each case we apply a new method, called Network-Guided Forests, to identify predictive modules together with logic functions which tie the activity of each module to the activity of its component genes. The resulting modules implement a diverse repertoire of decision logic which cannot be captured using the simple approximations suggested in previous work such as gene summation or subtraction. We show that in cancer, certain combinations of oncogenes and tumor suppressors exert competing forces on the system, suggesting that medical genetics should move beyond cataloguing individual cancer genes to cataloguing their combinatorial logic., Author Summary Biological outcomes are often determined by modules of proteins working in combination. In classic biological studies, these modules have been shown to encode a diverse repertoire of logic functions which provide the means to express complex regulatory programs using a limited number of proteins. Here, we integrate gene expression profiles and physical protein interaction maps to provide a systematic and global view of combinatorial network modules underlying representative developmental and cancer programs. We develop a new method that associates decision trees with concise network regions to identify network decision modules predictive of biological or clinical outcome. The resulting network signatures prove robust across different sample cohorts and capture causal mechanisms of development or disease. Furthermore, we find that the most predictive network decision functions rely on both coherent and opposing gene activities. Notably, in cancer progression the predictive gene associations often map to physical interactions between known oncogenes and tumor suppressors, where the combined activity of these genes determines disease outcome.

DNA Damage Pathways Revealed Despite the dynamic nature of cellular responses, the genetic networks that govern these responses have been mapped primarily as static snapshots. Bandyopadhyay et al. (p. 1385 ; see the Perspective by Friedman and Schuldiner ) report a comparison of large genetic interactomes measured among all yeast kinases, phosphatases, and transcription factors, as the cell responded to DNA damage. The interactomes revealed were highly dynamic structures that changed dramatically with changing conditions. These dynamic interactions reveal genetic relationships that can be more effective than classical “static” interactions (for example, synthetic lethals and epistasis maps) in identifying pathways of interest.

Background Recent development of mass spectrometry technology enabled the analysis of complex peptide mixtures. A lot of effort is currently devoted to the identification of biomarkers in human body fluids like serum or plasma, based on which new diagnostic tests for different diseases could be constructed. Various biomarker selection procedures have been exploited in recent studies. It has been noted that they often lead to different biomarker lists and as a consequence, the patient classification may also vary. Results Here we propose a new approach to the biomarker selection problem: to apply several competing feature ranking procedures and compute a consensus list of features based on their outcomes. We validate our methods on two proteomic datasets for the diagnosis of ovarian and prostate cancer. Conclusion The proposed methodology can improve the classification results and at the same time provide a unified biomarker list for further biological examinations and interpretation.

We have developed an integrated tool for statistical analysis of large-scale LC-MS profiles of complex protein mixtures comprising a set of procedures for data processing, selection of biomarkers used in early diagnostic and classification of patients based on their peptide mass fingerprints. Here, a novel boosting technique is proposed, which is embedded in our framework for MS data analysis. Our boosting scheme is based on Hannan-consistent game playing strategies. We analyze boosting from a game-theoretic perspective and define a new class of boosting algorithms called H-boosting methods. In the experimental part of this work we apply the new classifier together with classical and state-of-the-art algorithms to classify ovarian cancer and cystic fibrosis patients based on peptide mass spectra. The methods developed here provide automatic, general, and efficient means for processing of large scale LC-MS datasets. Good classification results suggest that our approach is able to uncover valuable information to support medical diagnosis.

Summary: Interrogating protein complexes and pathways in an evolutionary context provides insights into the formation of the basic functional components of the cell. We developed two independent Cytoscape plugins that can be cooperatively used to map evolving protein interaction networks at the module level. The APCluster plugin implements a recent affinity propagation (AP) algorithm for graph clustering and can be applied to decompose networks into coherent modules. The NetworkEvolution plugin provides the capability to visualize selected modules in consecutive evolutionary stages. Availability: The plugins, input data and usage scenarios are freely available from the project web site: http://bioputer.mimuw.edu.pl/modevo. The plugins are also available from the Cytoscape plugin repository. Contact: janusz@ucsd.edu Supplementary information: Supplementary data are available at Bioinformatics online.