First International Conference in Systems and Network Medicine: Applications of Systems Science and Thinking to Biomedicine Georgetown University Medical Center Washington DC,USA September 11–13,2019

Multi-Omics Network Construction and Its Analysis for Drug Repurposing

Seung-Hyun Jin ^* , YoungWoo Pae

MEDIRITA, INC., Republic of Korea

Objectives: We aim to introduce a method for multi-omics network construction and its analysis for drug repurposing.

Methods: Firstly, in order to construct the multi-omics network, we used the publicly available 28 databases ranging from genomics and metabolomics to disease and drug. From those databases, 12 categories of nodes and 18 categories of edges were extracted and systematically organized. Consequently, we constructed the multi-omics network composed of 85,040 nodes and 3,057,838 edges. Secondly, the breath-first search algorithm was applied when extracting nodes and edges from the multi-omics network. Finally, to refine and analyze the network obtained from the previous step, we applied network analysis measures such as clustering coefficient, betweenness centrality, modularity, and closeness.

Results: For example, we were able to check that ‘atorvastatin’ is related to breast cancer, coronary artery disease, and type 2 diabetes mellitus with high significance through our approach. Atorvastatin is known as a drug for the treatment of several types of dyslipidemias, and its relations to other disease such as breast cancer, coronary artery disease, and type 2 diabetes mellitus have also been reported. Thus, our example of atorvastatin supports the possible relations between atorvastatin and other diseases other than the current indication, from the multi-omics network aspect.

Conclusions: Multi-omics network constructed from public omics databases and the following network analysis might be useful for drug repurposing.

Domainscope: Protein Domain Based Disease Connections

Alin Voskanian ^* , Maricel Kann

University of Maryland Baltimore, MD, USA

Current gene-centric analyses to study mutations in coding regions are limited by their inability to account for the functional modularity of the protein. Previous studies of the functional patterns of known human mutations that have been implicated in cancer have shown a significant tendency to cluster at protein domain positions, namely position-based domain hotspots. By leveraging previous work done to incorporate over 80,000 mutations from various sources along with their disease association, this method gives new mechanism to find disease relations. Using this methodology disease relationships were identified that may not seem biologically obvious. The method identified 3381165 disease connected by the presence of similar domain mutation many of which are in druggable genes. Discovery of such relationships is helpful in drug repurposing, medical treatment and knowledge sharing among researchers. Future work will involve the addition of EHR information to more correctly and granularly categorize disease pairings.

Systems Pharmacology-based Uncovering Wogonoside as a Novel Angiogenesis Inhibitor of Triple Negative Breast Cancer by Targeting Hedgehog Signaling

Jiansong Fang ^* , Ofer Reizes, Justin Lathia, Charis Eng, Feixiong Cheng

Lerner Research Institute, Cleveland Clinic, OH, USA

Triple-negative breast cancer (TNBC) is an aggressive and heterogeneous disease that lacks clinically actionable genetic alterations that limit targeted therapies. Here we apply a systems pharmacology-based methodology, to identify novel, and consistent, therapeutic agents for treatment of TNBC via integrating drug-target networks and large-scale genomic profiles of TNBC. We identify wogonoside, one of the major active flavonoids, as a potent angiogenesis inhibitor. We validate that wogonoside attenuates cell migration, tube formation, and rat aorta microvessel outgrowth, and reduces formation of blood vessels in chicken chorioallantoic membrane and TNBC cell-induced matrigel plugs. In addition, wogonoside inhibits growth and angiogenesis in TNBC-cell xenograft models. The tissue-specific human protein-protein interactome network-based approach predicts, and we have empirically validate, wogonoside's anti-angiogenic effects as a consequence of VEGF secretion. Mechanistically, wogonoside inhibits Gli1 nuclear translocation and transcriptional activities of Hedgehog signaling in vitro and in vivo. Specifically, wogonoside inhibits the nuclear translocation and transcription activity of Gli1 by binding to SMO directly and specifically promoting ubiquitination-dependent degradation of SMO in TNBC cells. This study offers a powerful, integrated systems pharmacology-based strategy for oncological drug discovery and identifies wogonoside as a novel TNBC angiogenesis inhibitor.

Evaluation of Vanillin as a Potential Inhibitor of Proteins Involved in Signaling Pathway

Lesitha Jeeva Kumari^*1, Chellam Jaynthy¹, J., S. Usha²

¹Department of Bioinformatics, Sathyabama Institute of Science & Technology, Chennai, TamilNadu, India

²Department of Chemistry, Sri Sairam Engineering College, Chennai, TamilNadu, India

Cancer is one of the serious threats of human, which causes death. Worldwide, more than 20 million people are affected by cancer. Signaling cascades are the prime factors for cancer metastasis and therefore most anti-cancer treatments are targeted towards molecules of signaling pathway. There are about 13 signaling pathways in which each signaling pathway has it signaling network to transfer from one cell to the other. This helps to control the function of one or more cells that works together in a molecule. Therefore, in this study, we aimed at identifying the anticancer ability of a chemically synthesized lead compound para benzoyl vanillin semicarbazone against protein targets of the signaling pathways using molecular docking studies. Further we fragmented the compound and predicted the binding affinity of each of the fragments to identify the functional moiety of the compound. We also assessed the potential of the lead to act as a pharmacophore for designing chemotherapeutic ligands with better cytotoxic ability. Through our in silico intermolecular interaction studies we were able to narrow down to two specific targets for the lead namely MAPK and FRAT1. Fragmentation and molecular docking studies of the lead revealed that the bi benzyl group had a good binding affinity than the other fragments. Analogs of the bibenzyl fragment showed pharmacophore features namely a donor, two acceptors and a hydrophobe as the key functional unit and we postulate that these analogs derivatives can serve better in inhibiting the signaling pathway protein MAPK and FRAT1.

Models for Enhanced Cell-of-Origin Deconvolution of Cell-Free DNA

Megan E. Barefoot^*1, Marcel O. Schmidt¹, Rency S. Varghese¹, Yuan Zhou¹, Michael Lindberg¹, Yifan Chen¹, Sarah Martinez Roth¹, Loretta Yun-Tien Lin¹, Henghong Li¹, Alexander H. Kroemer², Habtom Ressom¹, Anton Wellstein¹

¹Department of Oncology, Lombardi Comprehensive Cancer Center, Georgetown University, Washington, DC, USA

²MedStar Georgetown University Hospital, Washington, DC, USA

Cell-free DNA (cfDNA) released from dying cells into the circulation is a potential indicator of tissue damage that could improve detection of treatment-related adverse events. As a non-invasive monitoring approach, changes in tissue-specific methylated DNA shed into the circulation may reflect intervention-based changes. Computational models aimed at deconvoluting cfDNA to trace cell turnover back to tissue- or cell-of-origin has become a significant research focus. In comparison to traditional analysis assessing methylation levels at individual CpG sites, pattern analysis assessing methylation levels across multiple adjacent CpG sites can increase sensitivity and specificity of signal detection. In this study, we planned to identify cfDNA methylation markers of heart, lung, and liver damage in clinical samples from patients and in mouse models. We describe a method to classify the tissue-origins of cfDNA at the level of single DNA molecules to explore a potential relationship between tissue-specific cell death and organ damage. As a result, we identify significant elevations of heart and lung cell-derived cfDNA in mice that received 8 Gy radiation compared to 3 Gy and sham control. In addition, we compare changing levels of liver cell DNA in patients at the time of pre-reperfusion relative to baseline and during serial serum samples collected shortly after transplant to assess the prognostic value of these markers. Our findings suggest that liver, heart, and lung derived cfDNA may be useful in monitoring treatment-related adverse events incurred through multiple forms of tissue injury. This approach can provide a better understanding of cell turnover in both health and diseased states and has vast applications to many areas of medicine studying tissue injury.

Metabolomics, Lipidomics, and Volatilomics Biobanking in Precision Systems and Network Medicine

Sinem Nalbantoglu^*1, Dilek Ceker¹, Yusuf Tambag², Hatice Aslanoglu³, Nilufer Kara³, Nilgun Algan³, Serdar Evman³, Volkan Baysungur^3,4, Abdullah Karadag¹

¹TUBITAK Marmara Research Center, Genetic Engineering and Biotechnology Institute, Kocaeli, Turkey

²TUBITAK BILGEM, Software Technologies Research Institute, Ankara, Turkey

³Ministry of Health, Istanbul Sureyyapasa Chest Diseases and Thoracic Surgery Research and Training Hosp., Istanbul, Turkey

⁴University of Health Sciences, Department of Thoracic Surgery, Istanbul, Turkey

Metabolomics involving metabonomics, lipidomics, and volatilomics are high-throughput analysis of the metabolome within biofluids, tissues to comprehensively identify and quantify all endogenous and exogenous low-molecular-weight (<1 kDa) small molecules/volatile organic compounds (VOCs). Metabolomics applications refer to a significant area in P-medicine, single cell metabolomics, epidemiologic population studies in accordance with phenomics, exposomics, DNA and protein adductomics, metabolic phenotyping/metabotypes profiling through metabolome-wide association studies (MWAS), and combination with other omics disciplines as integrative multi-omics. Systems Biology integrated omics technologies have been dedicated to comprehensively explain whole molecular biosignature of health and disease through P-medicine. In this aspect, accurate determination and validation of disease related biomarkers necessitates development of biorepository systems with a large collection and storage of patients' biospecimens through well annotated clinical and pathological data. Metabolomics biobanking eliminate or minimize variability, methodological challenges, systematic errors and bias, and enable analytical sensitivity and specificity, robustness, reproducibility, accurate quantitation, increased metabolite coverage with low metabolite losses, and qualified biomarker discovery. Here we provide discovery and validation findings of tissues and biofluids of a population of lung cancer patients using standardized and validated protocols of biobanking/biorepositories including well-established sample collection, storage, handling, and pre-analytical processing for untargeted and targeted metabolomics/metabonomics and volatilomics analyses. Our results highlight molecular phenotyping, patient stratification, and metabotypes with associated pathway networks of lung cancer patients in view of well-defined biobanking conditions which will further aid generation of individualized and targeted treatment options.

PTM Knowledge Networks and LINCS Multi-Omics Data for Kinase Inhibitor Drug Analytics in Lung Cancer

Karen E Ross^*1, Xu Zhang¹, Tapan K. Maity¹, Jake Jaffe², Cathy H Wu³, Udayan Guha¹

¹Thoracic and Gastrointestinal Oncology Branch, Center for Cancer Research, NCI, NIH, Bethesda, MD, USA

²Broad Institute of Harvard and MIT, Cambridge, MA, USA

³University of Delaware, Newark, DE, USA

Kinase domain mutations EGFR are common drivers of lung adenocarcinoma. Several generations of EGFR tyrosine kinase inhibitors (TKIs) have been developed to treat EGFR-driven lung cancers. Patients often have a good initial response to these drugs, but resistance inevitably develops, due to either additional EGFR mutations or to activation of parallel signaling pathways. To understand the mechanisms of resistance to the third generation EGFR TKI rociletinib, we conducted a mass spectrometry-based phosphoproteomic analysis comparing rociletinib-resistant and rociletinib-sensitive lung cancer cells. Using iPTMnet, a PTM resource that integrates data from text mining of the scientific literature and other PTM databases, we found that AKT and PKA kinases targeted many of the sites whose phosphorylation was up-regulated in resistant cells; these kinases may be part of signaling pathways that are aberrantly activated in these cells. Next, we used kinase-inhibitor target data (KinomeScan) and phosphoproteomic data (P100) generated by the NIH LINCS Program to identify drugs that might overcome drug resistance. Our study demonstrated that PTM knowledge networks can be used in conjunction with phosphoproteomic data to identify aberrantly regulated kinase signaling pathways in drug resistant cells, and that LINCS data (KinomeScan and P100) can be used to identify candidate drugs to be used in combination therapy to overcome resistance. Currently, we are testing drugs identified by LINCS analysis in cell culture assays, extending the analysis to other TKIs, and automating our workflow for overlay of PTM knowledge maps, LINCS data, and cancer omics data.

Utilizing Protein Association Data to Identify Unique Molecular Links between Diseases

Karan Luthria ^* , Olivier Bodenreider, Maricel G. Kann

University of Maryland Baltimore County, MD, USA

In recent years, the concept of drug repurposing has grown increasingly attractive due to the rising cost (>$2.5 billion) for new drug development. Current disease network models used to identify potential drug repurposing targets link diseases are based on shared genetic variants. However, genes and gene products commonly interact with other molecules that may contribute to different diseases. Therefore, there is a pressing need to include gene and protein interactions in current models for identifying disease relatedness and potential drug repurposing targets. Here, we combined gene-disease associated data with protein interaction networks to develop an extensive human disease-disease interaction network (HDDN) that can identify similarities between diseases at a complex molecular level. We compiled over 80,000 disease-variant associations from various manually curated databases. Since each database contains a unique disease descriptor, we leveraged the Unified Medical Language System (UMLS) to normalize the different terminologies. From the over 4000 genes associated with disease phenotypes, 2067 protein interactions were identified linking 28% of these genes. These protein interactions allowed for the development of the HDDN. When comparing our HDDN with the disease networks based on common genetic variants, we found 147% more molecular links between phenotypes. We have significantly enhanced existing disease network graphs through analyzing protein interactions. The HDDN provides an essential view on biological processes by revealing molecular links which are critical for effective hypothesis generation of drug repurposing targets and reducing healthcare costs.

Linking Advanced Multiplex Immunohistochemistry with Patient Clinicopathologic and Outcome Data to Facilitate Translational Research in Invasive Lobular Breast Cancer

Chaldekas KM^*1,2,6, Berry VL^1,3, Duttargi A¹, Harris BT^1,4,5, Riggins RB ⁶, Berry DL^1,6

¹Histopathology and Tissue Shared Resource, Lombardi Comprehensive Cancer Center, Georgetown University Medical Center, Washington DC, USA

²Department of Biochemistry and Molecular & Cellular Biology, Systems Medicine MS Program, Georgetown University, Washington DC, USA

³College of Arts and Sciences, University of Virginia, VA, USA

⁴Department of Pathology, Georgetown University Medical Center, Washington DC, USA

⁵Department of Neurology, Georgetown University Medical Center 6 Department of Oncology, Lombardi Comprehensive Cancer Center, Georgetown University Medical Center, Washington DC, USA

⁶Department of Oncology, Lombardi Comprehensive Cancer Center, Georgetown University Medical Center, Washington DC, USA

GUMC/LCCC's Histopathology and Tissue Shared Resource (HTSR) is developing specialized resources to associate patient biospecimen, clinicopathologic information, and advanced imaging and protein expression data to help researchers correlate biomarkers of interest with patient clinical outcomes. Reliably connecting data from different sources is critical for extracting meaningful output; however, translational researchers face persistent challenges as the amount of information that can be generated continues to grow. To expedite biomarker analysis, a tissue microarray (TMA) was constructed from tumor samples from 72 patients diagnosed with invasive lobular carcinoma (ILC). Multiplex immunohistochemistry (mIHC) was performed and TMA slides were digitally imaged on HTSR's Perkin Elmer Vectra-3 Multispectral Imaging System. The initial mIHC study generated 31 GB of data. Although Vectra datasets, the ILC TMA, and HTSR's breast cancer REDCap database are independently useful resources, their use is limited if they are not robustly linked. To allow investigators to explore novel ILC research questions and facilitate data sharing, HTSR's REDCap database was adapted and a semi-automated workflow developed to connect individual TMA samples with mIHC data to specific patients and their clinicopathologic information. Consequently, process metadata can be bulk uploaded into REDCap with each new mIHC study completed. Staining method, scoring approach, intensity thresholds, and image analyses such as phenotyping, tissue segmentation, and subcellular localization can be queried with relevant clinical data. Ultimately, the expanded database and workflow will be valuable foundational tools to link future Vectra datasets to diverse patient cohorts represented across HTSR-constructed TMAs.

Identification of potential immune targets in Endometrioid Endometrial Carcinoma metastatic progression using Ingenuity Pathway analysis

Eric Seiser ^* , Jean-Noël Billaud, Stuart Tugendreich, Debra Toburen

QIAGEN, Braintree, MA, USA

Endometrial adenocarcinoma is a common cause of gynecological cancer death in Europe and North America. The most dominant subtype, Endometrioid Endometrial Cancer (EEC) accounts for >80% of this cancerand is estrogen-dependent. At diagnosis, 75% of women have the disease confined to the uterus, which is considered Stage One. Five-year survival for Stage One patients is 80%, however, about 15–20% develop metastasis.

Total RNA extracted from tissues obtained after surgical resection from three women at stage one EEC was subjected to RNA-sequencing. The publicly available dataset (SRP045645) was downloaded directly from the Sequence Read Archive and the FASTQ files were processed with Biomedical Genomics Workbench for secondary analysis including mapping, quantification and differential expression analysis. Through streamlined integration the data was uploaded to Ingenuity Pathway Analysis (IPA) for biological interpretation.

We show how in silico solutions developed by QIAGEN Bioinformatics enable us to analyze the biological parameters involved in EEC metastatic progression from an early stage in the three patients diagnosed at Stage One. By comparing these patients' RNA-seq results, we determined that key Canonical Pathways and other biological processes differentiated the three patients from one other. In particular, the predicted transcriptional program allowed us to visualize key upstream drivers. Three of these transcriptional drivers are immune related molecules (two cytokines: CXCL14 and GDF15 and one growth factor FGF3). They have been predicted to be potential master regulators of a Causal Network that drives increase of EEC, metastasis, epithelial-to-mesenchymal-transition (EMT), and cellular invasion in one of the three patients. Based on these results we propose these three immune molecules as possible therapeutic targets to counteract the metastatic processes in EEC.

Omic Technologies and Frailty in Older Adults: State of the Science

Emma Kurnat-Thoma ^*

National Institutes of Health, National Institute of Nursing Research (NIH/NINR), Bethesda, MD, USA

Introduction: Frailty syndrome is a chronic and progressive condition primarily affecting adults >65 years old and increases risks for falls, hospitalizations, long-term nursing care and mortality. It is characterized by presence of at least three of five symptoms: slowness, weakness, exhaustion, inactivity and involuntary weight loss.

Problem Statement: Early diagnosis of frailty can assist health care providers and patients to optimize the highest possible functional status. However, frailty is a complex phenotype lacking a gold-standard diagnosis, is without scientific consensus on how to best operationalize it and is not routinely examined in many health care settings.

Methods: A focused review of the frailty scientific literature was conducted. Recent advances, translation opportunities, and implementation challenges across three perspectives were examined:1) clinical operationalization of the frailty phenotype in diagnosis and management; 2) biologic -omics Precision Medicine elements such as genomics, proteomics, epigenetics, microbiome and wearable sensor applications; 3) integrated care delivery frameworks and policy models able to support high-quality care at sustainable cost/value thresholds for frail individuals and populations.

Conclusion: Computational science provides a uniquely powerful opportunity to examine frailty across clinical, biological, and health care system components. Absence of scientific consensus on frailty's clinical operationalization results in a particularly severe translation gap for most applications.

Recommendation: Efforts to achieve U.S. scientific consensus on a frailty clinical diagnostic standard should be prioritized. Existing value-based structures, such as Accountable Care Organizations and Patient-Centered Medical Homes, are attractive policy models for piloting scalable older adult Precision Medicine interests.

Informatics Inference of Exercise Induced Modulation of Brain Pathways Based on Cerebrospinal Fluid Micro-RNAs (miRNA) in Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS)

Vaishnavi Narayan ^* , Narayan Shivapurkar, James N Baraniuk

Department of Medicine, Division of Rheumatology, Immunology and Allergy-Department of Medicine,

Georgetown University Medical Center, Washington DC, USA

Background: Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) subjects have pain, cognitive dysfunction, fatigue and exertional exhaustion after exercise.

Purpose: The post-exertional malaise of the subjects was modelled by comparing micro-RNA (miRNA) in cerebrospinal fluid from subjects who had no exercise versus submaximal exercise.

Methods: Differentially expressed miRNAs were examined by informatics methods to predict potential targets and regulatory pathways affected by exercise.

Results: miR-608, miR-328, miR-200a-5p, miR-93-3p and miR-92a-3p had higher levels in subjects who rested overnight (nonexercise n = 45) compared to subjects who had exercise before their lumbar punctures (n = 15). The combination was examined in DIANA mirPath v3.0, TarBase, Cytoscape and Ingenuity software^® to select the intersection of pathways and target mRNAs. DIANA found 33 targets that may be elevated after exercise including TGFBR1, IGFR1, and CDC42. Adhesion and Adherens junctions were the most frequent pathways. Ingenuity selected 7 targets that had complementary mechanistic pathways involving GNAQ, adenylcyclase 3 (ADCY3), RAP1B and PIK3R3. Potential target cells expressing high levels of these genes included choroid plexus, neurons, and microglia.

Conclusion: The reduction of this combination of miRNAs in cerebrospinal fluid after exercise suggested upregulation of phosphoinositol signalling pathways and altered adhesion during the post-exertional malaise of ME/CFS.

Confirmation that Exercise Challenge in Gulf War Illness Reveals Two Autonomic Subgroups with Altered Brain Function

Stuart D. Washington ^* , Rakib U. Rayhan, Richard Garner, Destie Provenzano, Kristina Zajur, Florencia Martinez-Addiego, John W. VanMeter, James N. Baraniuk

Department of Medicine, Division of Rheumatology, Immunology and Allergy-Department of Medicine, Georgetown University Medical Center, Washington DC, USA

Gulf War Illness (GWI) affects 25–32% of veterans from the 1990–1991 Persian Gulf War. Post-exertional malaise with cognitive dysfunction, pain and fatigue following physical and/or mental effort is a defining feature of Gulf War Illness. We modeled post-exertional malaise by assessing changes in functional magnetic resonance imaging at 3T during an N-Back working memory task performed prior to a submaximal bicycle stress test and after an identical stress test 24 hrs later. Serial trends in postural changes in heart rate between supine and standing defined three subgroups of veterans with Gulf War Illness: Postural Orthostatic Tachycardia Syndrome (GWI-POTS, 15%, n = 11), Stress Test Associated Reversible Tachycardia (GWI-START, 31%, n = 23), and Stress Test Originated Phantom Perception (GWI-STOPP, no postural tachycardia, 54%, n = 46). Before exercise, there were no differences in blood oxygenation level dependent activity during the N-Back task between control (n = 31), GWI-START, GWI-STOPP, and GWI-POTS subgroups. Exercise had no effects on BOLD activation in controls. GWI-START had post-exertional deactivation of cerebellar dentate nucleus and vermis regions associated with working memory. GWI-STOPP had significant activation of the anterior supplementary motor area that may be a component of the anterior salience network. There was a trend for deactivation of the vermis in GWI-POTS after exercise. These patterns of cognitive dysfunction were apparent in Gulf War Illness only after the exercise stressor. Mechanisms linking the autonomic dysfunction of START and POTS to cerebellar activation, and STOPP to cortical sensorimotor alterations, remain unclear but may open new opportunities for understanding, diagnosing, and treating Gulf War Illness.

Informatic Analysis of 3 miRNAs Decreased in Neuroblastoma Cells by GWI Serum

Baby Anusha Satravada ^* , James N. Baraniuk

Department of Medicine, Division of Rheumatology, Immunology and Allergy-Department of Medicine, Georgetown University Medical Center, Washington DC, USA

Gulf war illness GWI is a combination of potentially disabled disorder that can be seen in veterans of 1991 Gulf war. It is also generally named as chronic multi symptom illness. As the name implies it is not diagnosed easily due to its multiple symptoms such as Fibromyalgia, post-traumatic stress disorder and Migraine are some of the symptoms which we are focusing majorly. In this study we are trying to develop diagnostic biomarkers. In this study neuroblastoma cells were cultured with GWI or control serum for 3 days period. Each day cytoxicity was measured by MTS assay. On plotting these profiles, 3 cytoxicity patterns were seen. The pattern called gwi1 showed an upward trend for cell growth compared to controls serum. gwi2 showed upward trend on day 1 and 2 and downward trend on day 3. gwi3 showed downward trend. On 3^rd day RNA was extracted from these cells and quantitative PCR was performed to find miRNA profiles for control, gwi2 and gwi3. Cycle threshold (Ct) was set to 33 and delta delta Ct calculated for gwi2 and gwi3 compared to control. Significant changes were defined by a cutoff >2.Gwi2 and gwi3 shared 3 miRNAs that went up and one was diminished. Gwi3 had three additional miRNAs that were diminished.

Informatics analysis was performed to discover protein or gene targets for the different combinations. Methods will include pathway analysis, protein protein interaction analysis, protein and tissue enrichment analysis using various bioinformatics software's and tools such as DIANA, CYTOSCAPE, STRING DATABASE, KEGG, PATHWAY STUDIO, REACTOME. By this informatics analysis we will learn about mechanisms of serum induced toxicity in GWI. We can also see whether there are any drugs for the inhibition.

Informatic Analysis of microRNA-720 in GWI Serum Cytotoxicity

Sushma Maddipatla ^* , James N. Baraniuk

Department of Medicine, Division of Rheumatology, Immunology and Allergy-Department of Medicine, Georgetown University Medical Center, Washington DC, USA

Gulf war illness-GWI is a disabling disorder affecting the veterans of 1990–1991 Persian Gulf-war. It is also named as chronic multi symptom illness implying it is not diagnosed easily due to its multiple symptoms such as Fatigue, cognition and pain. In this study we are trying to develop diagnostic biomarkers for which neuroblastoma cells were cultured with GWI or control serum for 3 days period. Each day cytoxicity was measured by MTS assay. On plotting these profiles, 3 cytoxicity patterns were seen. The pattern called gwi1 showed an upward trend for cell growth compared to controls serum. gwi2 showed upward trend on day 1 and 2 and downward trend on day 3. gwi3 showed downward trend. On day 3 total RNA was extracted and miRNA was quantified by the QPCR and DDCt method. Gwi2 and gwi3 shared 3 miRNAs that went up and one was diminished. Gwi3 had three additional miRNAs that were diminished. GWI serum diminished miR-720 in gwi2 and gwi3 but not gwi1. Informatics-analysis indicates miR-720 may be derived from tRNA and possibly a 110bp gene on chromosome 3. The miR-720 biogenesis was predicted using mFold tool and literature. Literature search found miR-720 regulates NANOG for pluripotent stem cells, GATA3 for cell differentiation, cadherin and catenin for adhesion as gene-targets. miR-720 activity in integration with other two miRNA combinations will identify the mechanism of Regulated-Cell-Death. Lastly, to confirm the postulated changes in signaling pathways in GWI veterans invivo. Determine the biochemical nature of the “ToxicHumour” in GWI serum.

Machine Learning Techniques to Predict Readmission in Patients After Colorectal Cancer Surgery

Jim Huang ^* , Weibo Chen, Chengxin Liu, Nawar Shara, Mohammed Bayasi

MedStar Health Research Institute, Hyattsville, MD, USA

Successfully predicting readmission at the time of discharge supports clinical decision making for better continuity of care and helps hospitals avoid financial losses. This study compared the performance of artificial neural network (ANN) and decision tree (DT) with logistic regression (LR) in predicting the 30-days readmission, length of stay (LOS) and hospital charges at readmission using data spanning 2010–2017 from a multihospital system in the Mid-Atlantic region of the United States. 2,183 patients undergoing colorectal cancer surgery and discharged alive were identified. ANN and DT with bootstrapping aggregation algorithm were compared with LR with stepwise model selection. Predictors included demographics, diagnoses, comorbidities, complications, and procedure types. Models were developed using the training and validation sets, and the predictive performance was assessed using the test set. Performance of the models was assessed using prediction accuracy rate, sensitivity, specificity, and area under the ROC.

Overall readmission in our study cohort was 13%. The median LOS and charges at readmission were 6 days and $24,860, respectively. ANN outperformed DT and LR in predicting 30-day readmission overall (prediction accuracy: 73.6%, 64.2%, 48.3%) and in predicting higher level of LOS (prediction accuracy: 72.4%, 63.2%, 43.2%) and charges (prediction accuracy: 73.6%, 62.1%, 69.4%) at readmission. The study showed that patient demographics and clinical characteristics available at the time of discharge can predict 30-day readmission. The ANN model outperformed with the highest prediction accuracy compared with DT and LR models.

Implementing Machine Learning Techniques with Traditional Survival Analysis Methods to Assess the Use of SLFN11 as a Biomarker for Screening in Pediatric Solid Tumor Research

Charmin Guy ^* , Natasha Sahr

Department of Biostatistics, St. Jude Children's Research Hospital, Memphis, TN, USA

SLFN11, a gene of the Schlafen family, may serve as a potential biomarker for screening in certain childhood cancers. The degree of SLFN11 positivity is measured using H-score, the intensity of staining for the nuclei of tumor samples. These markers were used to elucidate the relationship between SLFN11 expression on the overall survival (OS), event-free survival (EFS), recurrence, and progression of disease in 335 samples from tumors in children and xenografts. Osteosarcoma and non-rhabdomyosarcoma were of specific interest for the effect of SLFN11 positivity on outcome when considering their reduced apoptotic functioning resulting in a worse response to DNA damaging agents used in chemotherapy and radiation. All 7 diagnostic groups were examined using various methods of survival analysis and machine learning to highlight the importance of interpreting data in multiple forms. SLFN11 expression did not induce significant effects on the OS, EFS, recurrence, or progression of any of the groups in traditional statistical analysis but may be clinically relevant for screening based on machine learning classifications.

Early Prediction of Hemodynamic Shock in Intensive Care Unit Using Deep Learning and -Linear Time-Series Features

Aditya Nagori^*1, Pradeep Singh², Prakriti Ailavadi³, Shubham Yadav³, Rakesh Lodha², Tavpritesh Sethi^2,4

¹CSIR-Institute of Genomics and Integrative Biology, New Delhi, India

²All India Institute of Medical Sciences, New Delhi, India

³Netaji Subhas Institute of Technology, New Delhi, India

⁴Indraprastha Institute of Information Technology, Delhi, India

Hemodynamic shock is a rapidly killing condition in intensive care units (ICUs), emergency room and in the community settings. Shock often leads to compromise and failure of multiple organs due to decreased tissue function. Mortality rate in patients who develop shock in ICU is high as 34%. Timely monitoring is important to prevent any secondary organ dysfunction and mortality. Early recognition has been shown to be effective. Thus, it is highly desirable to have a system for shock detection and prediction. We in this work used routinely collected physiological vitals data of the patients to predict hemodynamic shock in ICU. We used MIMIC-III database to first create a cohort on multivariate time-series data, where 781 Non-shock and 1210 Shock cases were obtained after screening patients who developed shock in ICU. The dataset is splitted into train, validation and test set in 60%, 20%, 20% ratios. The training set is further oversampled for the minority class. We used two major approaches to build prediction models. First we built a four branched “Long short-term memory” (LSTM) model which takes input of 256 minute length of four physiological signals and predict the next 3 hour Shock status, this model achieved AUC of 83% on the test-set. Second we extracted non-linear time-series features which are passed to a feature selection pipeline “Boruta” to build a more parsimonious model, as a result 226 important features were obtained. Finally, random-forest model was built on these features which resulted into AUC of 84% on test-set.

A Machine Learning Approach for Disease Genes Signatures

Annalisa Longo², Venkata Pochiraju³, Daniele Santoni¹, Davide Vergni¹, Paolo Tieri^*1

¹CNR-IAC, Rome, Italy

²Tor Vergata University, Rome, Italy

³Sapienza University, Rome, Italy

In the context of network medicine, disease genes, i.e. genes that have been experimentally associated to the onset or progression of a pathology, show a complex set of features that are not easily reduced to, and grasped by a simple network approach (e.g., studying centrality measures or clustering characteristics of the gene network). Here, to overcome such limitations and to exploit a larger set of informational attributes available, we analyze a sizeable integrated set of biological, ontological and topological features (including interaction data and GO categories, among others) related to different collections of disease genes (including, but not limited to sets related to several inflammatory and dysmetabolic diseases) via a comprehensive machine learning (ML) approach, in order to discover recurring patterns of attributes associated to families of disease genes. In this way the chances of revealing complex, hidden topological, ontological and statistical properties of the genes under scrutiny is wider and the derived “signature” can be heuristically used in a discovery process to find further yet unknown disease genes. We show hurdles, discriminating capabilities and main results in sorting out and in reconstructing the feature sets, in selecting the appropriate ML approach and in analyzing the datasets.

Early Prediction of Childhood Asthma Exacerbations Through a Combination of Statistical and Machine Learning Approaches

Aditya Nagori^*1, Tavpritesh Sethi^2,3, Sushil Kumar Kabra², Rakesh Lodha², Anurag Agrawal¹

¹CSIR-Institute of Genomics and Integrative Biology, Delhi, India

²All India Institute of Medical Sciences, New Delhi, India

³Indraprastha Institute of Information Technology, Delhi, India

Severe exacerbations of asthma cause irreversible decline in lung function especially in growing phase of life. Thus, minimizing exacerbations is the primary goal of clinical management yet, integrative frameworks for driving better practice are lacking. We recorded and analyzed data on a five-year longitudinal pediatric cohort, n = 256 followed at quarter-yearly intervals with the objective of forecasting an impending exacerbation. Data included clinical, physiological, metabolomic, biochemical and environmental variables. We introduced derived features such as exacerbation scores, hypothesis driven variables, rate of growth, variability of lung function etc. Exploratory data analysis upon these 243 variables revealed complex interactions thus, motivating the application of Boruta, a machine learning algorithm particularly suited for complex structure. A set 50 and 56 variables for exacerbation scores and event prediction respectively found to be important, further used for model construction. Observations were partitioned into 10-fold Training (75%) and Test (25%) sets. Mixed effect regression models were trained on selected variables. Parsimonious models were selected for exacerbation event and scores prediction. In event prediction model, previous exacerbation scores, Impulse oscillometry (IOS) based impedance (5Hz), reactance (5Hz) were obtained as final predictors. Model on test-set achieved average AUC of ∼73%. While in the exacerbation scores prediction model, previous exacerbation scores, reactance (5Hz) found to be final predictors, achieved average R-square = 93% on test set. Thus, an integrative data-science approach using a combination of machine learning and classical statistical analysis can enable clinical decision making for complex scenarios such as forecasting of exacerbation in childhood asthma.

Systems Medicine Education at Georgetown University Medical Center: Program Development and Description of Collaborative Efforts with The Program Director and Librarians

Douglas Varner^*1, Sona Vasudevan²

¹Dahlgren Memorial Library (DML), Georgetown University Medical Center, Washington DC, USA

²Department of Biochemistry, Georgetown University Medical Center, Washington DC, USA

This poster describes the Systems Medicine Master's Program at the Georgetown University Medical Center (GUMC) including the unique history and origins of the program at GUMC and how the program has evolved through time to its current status as a stand-alone master's program in the Biomedical Graduate Education sector of GUMC. GUMC has a rich history in pioneering efforts with computer methodologies for coding and comparison of protein and nucleotide sequencing with the work of Dr. Margaret Dayhoff which led to the development the Protein Information Resource. Work by Dr. Dayhoff and other researchers at GUMC led to the development of the Systems Medicine Master's program in 2011. Initially offered to medical students as a combined MD/MS degree the program evolved in 2016 into a stand-alone master's program with 10–15 students per yearly cohort. The curriculum is designed to give the students a rich learning experience with didactic coursework and opportunities for research and internships drawing on the extensive expertise from Georgetown faculty and a community of highly specialized researchers from the Washington, DC area. This poster will also describe a unique collaboration between the program director, an active bioinformatics researcher and educator, and a health sciences librarian from the Dahlgren Memorial Library, the Graduate Health & Life Sciences Research Library at GUMC. This collaboration involved development of two journal club courses and additional touch-points in the curriculum as guest lecturer and instruction of students in the use of library and information resources for completion of their coursework and projects.

Do Institutional Orthopedic Trauma Databases Provide Accurate Information?

Aman Chopra ^* , Abigail C. Cortez, M.D., Ashraf El Naga, M.D., Anthony Ding, M.D., Saam Morshed, M.D., Ph.D.

University of California, San Francisco, CA, USA

Academic trauma institutions rely on fracture databases as a research and quality control tool. The purpose of this study is to determine the capture rate of a resident-populated database in collecting information on extremity fractures treated to determine the accuracy of resident-assigned Orthopaedic Trauma Association (OTA) fracture classifications. A retrospective study was performed at a trauma center of adult patients who underwent definitive treatment for extremity fractures. A random sample was taken from these entries and compared to a resident-populated fracture database designed to capture the same patients. For all matching records containing a resident-assigned OTA classification, relevant imaging was blindly reviewed by an orthopaedic trauma surgeon. Resident OTA classifications were compared to this gold standard to determine accuracy rate. 2002 unique fracture records were identified from the billing database between 2012 and 2017. A 20% sample of 400 entries was randomly selected. 318 (80%) out of 400 entries were captured by resident populated database. 231 of these 318 entries contained an OTA classification. 153 (66%) of these 231 entries demonstrated concordance between resident and attending assigned OTA classifications. 133 of the 190 lower extremity classifications were accurately identified as compared to 20 of the 41 upper extremity classifications (p = 0.009). 79 of the 121 end segment fractures showed agreement versus 42 of the 63 diaphyseal injury patterns (p = 0.85). Classification accuracy did not significantly vary by resident year of training (p = 0.142). Trainee generated databases may be subject to incomplete data entry and inaccurate fracture classifications.

Influence of WNT and DNA Damage Response Pathway Alterations on Outcomes in Patients with Unresectable Metastatic Colorectal Cancer

Sebastian Mondaca¹, Henry S Walch^*2, Subhiksha Nandakumar², Walid K. Chatila², Jaclyn Frances Hechtman³, Andrea Cercek¹, Luis A. Diaz¹, Francisco Sanchez-Vega², Nancy E. Kemeny¹, Neil Howard Segal¹, Zsofia Kinga Stadler¹, Anna M. Varghese¹, Efsevia Vakiani³, Marc Ladanyi⁴, Michael F. Berger⁵, David B. Solit⁶, Jinru Shia³, Leonard B. Saltz¹, Nikolaus D Schultz⁷, Rona Yaeger¹

¹Department of Medicine, Memorial Sloan Kettering Cancer Center, New York, NY, USA

²Marie-Josée and Henry R. Kravis Center for Molecular Oncology, Memorial Sloan Kettering Cancer Center, New York, NY, USA. et al. Full affiliation in the website

³Departments of Pathology, Memorial Sloan Kettering Cancer Center, New York, NY, USA

⁴Departments of Pathology, Memorial Sloan Kettering Cancer Center, New York, NY, USA; Human Oncology and Pathogenesis Program, Memorial Sloan Kettering Cancer Center, New York, NY, USA

⁵Marie-Josée and Henry R. Kravis Center for Molecular Oncology, Memorial Sloan Kettering Cancer Center, New York, NY, USA; Departments of Pathology, Memorial Sloan Kettering Cancer Center, New York, NY, USA; Human Oncology and Pathogenesis Program, Memorial Sloan Kettering Cancer Center, New York, NY, USA

⁶Department of Medicine, Memorial Sloan Kettering Cancer Center, New York, NY, USA; Marie-Josée and Henry R. Kravis Center for Molecular Oncology, Memorial Sloan Kettering Cancer Center, New York, NY, USA; Human Oncology and Pathogenesis Program, Memorial Sloan Kettering Cancer Center, New York, NY, USA

⁷Marie-Josée and Henry R. Kravis Center for Molecular Oncology, Memorial Sloan Kettering Cancer Center, New York, NY, USA; Departments of Epidemiology and Biostatistics, Memorial Sloan Kettering Cancer Center, New York, NY, USA; Human Oncology and Pathogenesis Program, Memorial Sloan Kettering Cancer Center, New York, NY, USA

Background: We assembled a large series of consecutive patients with unresectable metastatic colorectal cancer (mCRC) to identify genomic biomarkers of response and survival.

Methods: Patients with unresectable mCRC treated at Memorial Sloan Kettering with genomic tumor profiling between 2014 and 2017 were included. Patients who underwent upfront metastasectomy or received neoadjuvant/conversion chemotherapy were excluded. We evaluated associations between genomic profiles with progression free survival (PFS) on first-line chemotherapy and overall survival (OS).

Results: Of 1453 mCRCs profiled in this period, 471 patients met the study criteria. Most tumors (91%) were microsatellite stable (MSS). Among MSS patients treated with oxaliplatin-containing regimens (n = 305), 7% harbored alterations in genes associated with DNA damage response (DDR). DDR gene alterations were not associated with PFS (P = 0.94). Genomic alterations that significantly varied by duration of response included BRAF (16%, 10%, and 5% for PFS <6 months, 6–12 months, and >12 months, respectively) and APC (62%, 74%, and 80% for PFS <6 months, 6–12 months, and >12 months, respectively). APC mutation was associated with significantly longer PFS (HR 0.67) and OS (HR 0.59) in multivariate analysis versus no WNT pathway alteration or alterations in other WNT pathway genes.

Conclusions: In unresectable mCRC patients, mutations in APC were associated with better outcomes; absence of an APC alteration or the occurrence of other WNT pathway alterations was associated with shorter survival. Somatic alterations in DDR genes were not associated with outcomes in mCRC patients receiving oxaliplatin-containing regimen.

7P Systems Medicine: Incorporating Personhood into Systems Medicine

James A. Marcum ^*

Baylor University, TX, USA

A major criticism of 4P systems medicine is that the holism its advocates espouse is not the traditional humanistic holism but a technical one. The patient's personal preferences and values are excluded from the omics data clouds; and, the patient qua person is reduced to a calculated or quantified entity. To address this criticism, a 7P medicine is proposed that includes the personhood of patients. To assimilate personhood into systems medicine, the proposed model is constructed from the P-components, consisting of a physico-physiological P, a psycho-cognitive P, and a public-populational P. These three P-components are then integrated in terms of a philosophical notion of personhood to engender a fourth P—a personal P. Each of these P-components is implemented clinically through three P-operations composed of a predictive P, a participatory P, and a preventive P. In executing 7P medicine, the clinician utilizes omics data and personal preferences and values from each of the P-components to predict a possible health risk and then to solicit the patient's participation with respect to preventing the risk and thereby promoting health. The poster concludes with challenges facing 7P medicine in terms of implementing it clinically.

Local Network Topology Differences Between Early and Late Recurrence in ER+ Breast Cancers

Robert Clarke, Mike Dixon, Lu Jin ^* , Pearce Dominic, Arran Turnbull, Cigdem Selli, Rong Hu, Alan Zwart, Yue Wang, Jason Xuan, Surojeet Sengupta, Andy Sims, Minetta C. Liu

Lombardi Comprehensive Cancer Center, Georgetown University Medical Center, Washington DC, USA

Background: Many breast cancers recur years after their initial endocrine treatment. If systemic endocrine therapies killed all cells, recurrence would reflect only the appearance of new disease. Thus, we hypothesized that cells that survive and lie dormant may be driven, in part, by altered wiring of their cell death signaling. We, therefore, studied how cell death signaling is differentially wired in primary tumors that will recur early versus those that will recur later.

Method: Genes involved in cell death related pathways were identified from KEGG to initiate our Knowledge-fused Differential Dependency Network (KDDN) analysis of gene expression data from public and our in-house gene expression dataset. Common gene-gene interactions (edges) predicted in at least two different datasets were extracted from all KDDN analyses results. To strengthen the relevance of these features, predicted network edges that represent known protein-protein interactions (PPI) were identified from the STRING database, and these edges were noted in the signaling graphs. Final network graphs were constructed using the common edges from all overlaid networks.

Results: MAPK8 and CYCS (Molecular Mechanisms of Cancer, p = 1.58E-52), TNFRSF1A (Neuroinflammation Signaling Pathway, p = 1.26E-54), RELA, and NFKB1 (Colorectal Cancer Metastasis Signaling, p = 7.94E-35), were identified as hubs. Hubs may be critical signaling components driving the differences between tumors that will become dormant and recur late. With longer follow-up time, we will better define the two groups and identify additional common phenotype specific gene-gene interactions.

Critical Nodes Discovery in Pathophysiological Signaling Pathways

Enrico Mastrostefano, Marco Cianfriglia ^* , Alessandro Celestini, Paolo Tieri

CNR-IAC, Rome, Italy

Network-based ranking methods (e.g. centrality analysis) have found extensive use in systems medicine for the prediction of essential proteins, for the prioritization of drug targets candidates in the treatment of several pathologies and in biomarker discovery, and for human disease genes identification. Here we propose to use critical nodes as defined by the Critical Node Problem for the analysis of key physiological and pathophysiological signaling pathways, represented as directed networks, as target candidates for treatment and management of several cancer types, neurologic and inflammatory dysfunctions, among others. The Critical Node Problem has been proven to be NP-hard, thus in our study we use a heuristic for directed graphs, that was recently proposed and proved performing better than previous heuristics, to detect critical nodes. We also implemented a brute-force procedure to compare heuristic's results when the graph is small enough to apply it. We show how critical nodes allow to rank the importance of proteins in the pathways in a non-trivial way, substantially different from classical centrality measures and highlighting the set of critical nodes whose deletion maximize the disconnection of the graph. Such ranking takes into account the extent to which the network depends on its key players to maintain its cohesiveness and consistency, and coherently maps biologically relevant characteristics that can be critical in disease onset and treatments. Such approach can provide valuable support in the process of key player alteration identification as well as in drug target and therapeutic options management.

Investigating the Association Between HbA1c Levels and Stroke Incidence Using HER Data: Personalized Medicine Approach

Kanchan Bhasin^*1, Belqes Ahmadi¹, Nawar Shara², Sona Vasudevan¹

¹Department of Biochemistry, Molecular and Cellular Biology, Georgetown University, Washington DC

²MedStar Health Research Institute, Hyattsville, MD, USA

It is a well-established fact that diabetes mellitus is one of the major risk factors for ischemic stroke and previous studies have shown that there is a possible association between the HbA1c levels and risk of stroke incidence in both diabetic as well as non-diabetic population. As stroke is one of the top leading causes of death and morbidity worldwide, it is important to focus on newer and preventative methods which could lead to early detection. Therefore, this study focused on validating the correlation between HbA1c levels and stroke incidence from the EHR patient data by using the statistical analysis tool ‘R’. Multiple variables such as age, gender, ethnicity, height, weight, systolic and diastolic blood pressure, smoking history, history of diabetes, HbA1c levels recorded pre- and post-Stroke event were considered to build the statistical models with an aim to identify the best predictive model for the association. The regression analysis, conducted on 1,902 samples from the MedStar patient database, showed that there is a strong correlation between diabetes and stroke. Three out of the four regression models had ‘diabetes’ as a significant variable in the final models suggesting that regular assessment of HbA1c levels in diabetic patients can prove to be an important predictor of stroke risk. A rise in the HbA1c levels should indicate increased risk for stroke in such patients. However, future studies need to be conducted in non-stroke and non-diabetic population as well to better understand and validate the association of HbA1c levels to stroke incidence.

Analysis of Host Gene Expression in HEK-293 Cells Infected with Zika Virus

Navin Vijayarangan^*1, Tadahisa Teramoto², Radhakrishnan Padmanabhan²

¹Departments of Biochemistry & Molecular and Cellular Biology, Georgetown University, Washington, DC

²Microbiology and Immunology, Georgetown University, Washington, DC

Zika Virus (ZIKV), newly emerged mosquito-borne human pathogen, affects millions of people around the world. Proper genetic understanding of virus-host interaction is required to assess the pathogenicity of the virus. ZIKV infection of a human cell line, HEK-293 upregulated as well as downregulated a number of host genes as revealed by RNAseq analysis. In this study, we selected a set of four upregulated genes based on their fold change and P values and validated them by quantitative real time PCR (qRT-PCR). All the selected genes were successfully validated and can be candidates for metabolic analysis.

The Investigation of Diabetes Remission by Bariatric Surgery Metabolism

Shan Zaidi ^* , Ancha Baranova

George Mason University, Fairfax, VA, USA

Bariatric surgery, primarily used to treat obesity, has shown to elicit a remission in type 2 diabetes mellitus in 40 - 50% of patients. After analysis of existing literature concerning metabolic control exerted by gut hormones, we hypothesized that oxyntomodulin (OXM), peptide YY (PYY), pancreatic polypeptide (PPY), ghrelin, and glucagon-like peptide 1 (GLP-1) contribute to post-surgery anti-diabetic effects. To find support for this hypothesis, analysis of diabetes-associated genetic networks was performed in Pathway Studio environment. Application of OXM was shown to inhibit ghrelin release, thus, displaying a potential for a development as novel therapeutic. Lowered concentrations of PYY in plasma may contribute to both type 2 diabetes and obesity. The list of molecules with levels elevated post-surgery include GLP-1 and agonists for its receptors, well-known to improve insulin and normalize glucagon defects in diabetic patients. Comprehensive understanding of the post-bariatric changes in gut hormones milieu may produce novel targets for the development of novel anti-diabetic medications.

Informatics View of Inflammatory Bowel Disease, Crohn's Disease and Ulcerative Colitis

Takashi Kitani ^* , Sona Vasudevan

Georgetown University Medical Center, Georgetown University, Washington, DC, USA

Crohn's Disease (CD) and Ulcerative Colitis (UC), both under the umbrella of Inflammatory Bowel Diseases (IBD), involve distinct molecular processes. The difference in molecular processes can be studied using different gene members involved in each of the diseases, to find potential drug targets as well as other clinically relevant details.

Previously, Dr. Vasudevan's group developed a manually curated database of genes, SNP, miRNA involved in IBD as well as a few other autoimmune diseases. The genes identified were analyzed using analytics software as IPA and Pathway Studio, to reveal the molecular processes involved in each of the diseases.

We identified that CD involved eosinophils and cellular immunity while UC involved Th17 and humoral immunity. UC seem to involve interactions with external environment for its disease state. These molecular differences associate CD and UC to distinct inflammatory diseases and processes, and yield different drug targets and drug candidates. Further, we also identified that what is classified as ‘IBD’ have significant overlaps with CD and UC, and that they may be unclassified forms of CD and UC.

Using these data, we identified potential similar diseases to CD UC and IBD, and further identified drug candidates currently in use for other diseases.

We demonstrated that CD and UC are different diseases using informatics, and also showed similarities between the two diseases and ‘IBD’. These showed insight into disease characteristics that can help potentially with future diagnosis of these diseases as well as treatment.

A Longitudinal Study of Cancer Therapy-Related Cardiac Dysfunction by Network-Based Analysis of Large-Scale Echocardiograms

Yuan Hou^*1, Zhou¹, Muzna Hussain², Thomas Budd³, W. H. Wilson Tang⁴, Chirag Shah⁵, Rohit Moudgil⁴, Zoran Popovic², Brian Griffin², Mohamed Kanj⁴, Patrick Collier^2,6, Feixiong Cheng^1,6,7

¹Genomic Medicine Institute, Lerner Research Institute;

²Department of Cardiovascular Imaging,

³Department of Hematology/Medical Oncology;

⁴Robert and Suzanne Tomsich Department of Cardiovascular Medicine, Sydell and Arnold Miller Family Heart and Vascular Institute;

⁵Department of Radiation Oncology, Cleveland Clinic, Cleveland, OH, USA

⁶Department of Molecular Medicine, Cleveland Clinic Lerner College of Medicine, Case Western Reserve University, Cleveland, OH, USA

⁷Case Comprehensive Cancer Center, Case Western Reserve University School of Medicine, Cleveland, OH, USA

*Presenter/Corresponding author: Feixiong Cheng chengf@ccf.org

ABSTRACT NOT AVAILABLE FOR PRINT

Precision Systems and Network Medicine: Metabolomics Profiling of Prostate Cancer Clinical Characteristics Following Radiotherapy

Sinem Nalbantoglu^1,2, Mones Abu-Asab³, Simeng Suy, Sean Collins⁴, Hakima Amri^*1

¹Department of Biochemistry, Cellular and Molecular Biology, School of Medicine, Georgetown University, Washington DC, USA

²TUBITAK Marmara Research Center, Institute of Gene Engineering and Biotechnology, Molecular Oncology Laboratory, Gebze, Kocaeli, Turkey

³Section of Ultrastructural Biology, NEI/NIH, Bethesda, MD, USA

⁴Department of Radiation Oncology, School of Medicine, Georgetown University, Washington DC, USA

Systems biology integrated metabolomics profiling of prostate cancer (PCa) patients following radiotherapy (RT) would shed light on revealing treatment outcomes and developing precision biomarkers. We analyzed the correlation of parsimony phylogenetic systems biology integrated mass spectrometry-based untargeted serum metabolomics of PCa patients (n = 55) with their clinical parameters before and after treatment. The protocol consisted of stereotactic body radiation therapy (SBRT) and intensity modulated RT (IMRT) with SBRT. The RT generated 5 phylogenetic subgroups with distinct metabolomic profiles that did not correspond to hormonal treatment, risk assessment, metastasis, or PSA levels. PSA was neither a factor influencing clade membership nor an indicator of risk assessment or metastasis. Moreover, the hormone-treated patients didn't form their own clade but rather spread among the five clades. The same applies to risk assessment and metastasis. The cladogram offers a new outlook on the clinical variables that may not be the most significant indicators of health outcome. The 88 significantly altered pre-RT and 29 post-RT features showed aberrations in the metabolic pathways of nitrogen, pyrimidine, purine, porphyrin, alanine, aspartate, glutamate, and glycerophospholipid. Bilirubin, phthalic acid, 5’-Benzoylphosphoadenosine, and carbamic acid were altered in all of the metastatic post-RT patients. D-tryptophan, hypoxanthine, tetrahydroisoquinoline, dihydrosanguinarine, and methylglutaric acid were detected in pre-RT metastatic patients, and this may not be related to treatment or RT. In conclusion, those compounds should be further investigated for their association to metastasis, treatment response, side effects, and radiation outcomes. Individual metabolic profiles and associated clinical phenotypes may lead to precision treatments and improved health.

From Bugs to Diseases: An Analysis of Similarities in Microbiota Changes Across Diseases

Natasha Raja^*1, Mansi Maini^*1, Douglas Varner², Sona Vasudevan¹

¹Department of Biochemistry, Cellular and Molecular Biology, Systems Medicine Program, School of Medicine, Georgetown University, Washington DC, USA

²Dahlgren Memorial Library (DML), Georgetown University Medical Center, Washington DC, USA

The human microbiome includes trillions of symbiotic microbial cells including bacteria, archaea, viruses and fungi. These organisms play a significant role in regulating metabolic and immune functions. Recent literature has revealed growing evidence of the intricate role the microbiome plays in health and disease. Illnesses such as Obesity, Diabetes, Colorectal Cancer, and Asthma have been linked to changes in the human microbiota. Specifically, the variety and abundance of species found in an individual's gut, may be linked in many different disease pathologies. This study analyzes major human diseases in the context of organisms found in each of the diseases. More specifically, this study investigates if certain diseases share similar bacterial species, conducts a comparative analysis on their prevalence (high or low), and analyzes if these species share a common pathway.

Variability and Persistence: Better Clinical Practice with Better Systems Concepts

James Palmer ^*

Caldwell Palmer, Denver, CO, USA

Systems medicine concepts and thereby clinical practice are both being significantly improved by understanding how human beings exist paradoxically as both continuously dynamically variable and continuously dynamically persistent. The proposition is asserted that dynamic variability analysis in general outperforms more static point and threshold measures of the embodied person. Particular dynamic variability analytics and supporting theories for clinical practice including outcomes relevant to all involved, will be described. The use and increased range of dynamic analytics in practice, provides new light back on to medically relevant basic concepts of being a human (science ontology). The first variability example is mortality risk of COPD caused lung damage as assessed by fractal dimension structure analysis (a temporal-spatial scaling symmetry statistic). Other uses cited of fractal dimension or time series, other variability analytics are: blood clot structure; HTN drug class performance; infection/sepsis trajectories. Implications of math symmetries for human variability are discussed. The second example integrates psychological variability and physical and social interactions variability with outcomes effects for: Diabetes Type II risk in 5000 shipyard workers and Glycemic control levels in diabetic patients. The variability analytic used is Sense of Coherence psychometric – SOC13.