Abstract
Severe asthma fingerprints: from discovery to point of care
Severe asthma exhibits far from simple pathophysiology underlying a heterogenous clinical presentation. The complex biology of severe asthma requires the unravelling of various disease phenotypes. Even though single biomarkers (e.g. sputum eosinophils and exhaled nitric oxide) are gradually introduced in the management of severe asthma (based on the latest meta-analyses), the U-BIOPRED Study (http://www.ubiopred.eu) funded by IMI has postulated that composite biomarker fingerprints will provide more comprehensive patho-biological information of these diseases and their underlying molecular phenotypes.
The pediatric and adult cohorts of U-BIOPRED exhibit several clinical phenotypes as appeared from cluster analyses. High-throughput “omics” technologies based on unbiased systems biology approaches, including transcriptomics, proteomics, lipidomics, and metabolomics are being used by U-BIOPRED for biomarker discovery in severe asthma. The leading principle here is to strictly obey the recent guidelines on the validity of omics analysis in clinical medicine.
Even though gene expression profiles in blood, sputum and bronchial brushes and biopsies differ between clinical and inflammatory subgroups (eosinophilic, neutrophilic, Th2, non-Th2, IL-1 receptor) these are comprising complementary information leading to newly discovered severe asthma phenotypes. Notably, gene expression does not track between sample locations (blood, sputum, biopsies) and differs between adult- and childhood-onset asthma and between patients with and without fixed airflow limitation.
These data indicate that there are multiple bio-clinical phenotypes of severe asthma . Non-invasive technologies, such as real-time breathomics are bringing these fingerprints to point of care. This is facilitating data-driven management not only in asthma, but also in COPD and lung cancer.
Network-based disease classification - de novo and non-de novo endophenotyping in cancer
On major obstacle in current medicine and drug development is inherent in the way we define and approach diseases. Here, we will discuss the diagnostic and prognostic value of (multi-)omics panels in general. We will have a closer look at breast cancer subtyping and treatment outcome, as case example, using gene expression panels - and we will discuss the current “best practice” in the light of critical statistical considerations. Afterwards, we will introduce computational approaches for network-based medicine. We will discuss novel developments in graph-based machine learning using examples ranging from Huntington's disease mechanisms via lung cancer drug target discovery back to where we started, i.e. breast cancer subtyping and treatment optimization - but now from a systems medicine point of view. We conclude that systems medicine and modern artificial intelligence open new avenues to shape future medicine.
Hands-on tutorial on object-oriented modelling of large-scale metabolic models
Faculty of Computer and Information Science, University of Ljubljana,
Large-scale metabolic models in general and large-scale metabolic models of liver metabolism in specific are gaining large importance within the research community. Their first clinical applications have been reported in recent years in the context of personalised and systems medicine. Most of these models are based on the stoichiometric description of the metabolic reactions encoded in the genome of the observed organism. These models, however, do not account for the interactions of the liver with surrounding tissues and lack the direct integration of metabolic pathways with gene regulatory and signalling networks.
In our tutorial we will provide the hands-on demonstration of the alternative modelling approach based on object-oriented modelling. We will describe the in-house developed SysBio library that can be used to construct large-scale metabolic models. This library allows the user to build complex biological models simply by connecting its basic objects, and provides the means for the integration of metabolic, gene regulatory and signalling reactions. The library presumes the normalised steady-state description of the system's response, which allows us to drastically reduce the number of kinetic parameters needed to perform the simulations. We will demonstrate the application of the SysBio library on several examples. These will include the use of the SysBio-based SteatoNet model and its extension to the LiverSex model, i.e. the first sex-based multi-tissue and multi-level liver metabolic model that provides the insights into the sex-dependent complex liver pathologies.
We present Grohar, a computational tool focused to the analysis, visualization and alignment of genome scale metabolic models (GEMs). Grohar provides an easy to use graphical interface, which allows the user to perform different types of COBRA analyses without any programming skills. The analyses are mainly based on the evaluation of metabolic fluxes with flux balance analysis (FBA) and its derivations. User can manually select the segments of interests of metabolic networks, upon which the visualizations are performed. User can also visualize the effects of the perturbations performed upon the selected segments. Moreover, Grohar allows automatic identification of individual metabolic pathways within the GEM network. Metabolic pathways in the form of SBML models or KEGG maps are aligned with the GEM model. This allows the user to automatically identify the given pathway within the larger model. The alignment is performed with the combination of different algorithms that evaluate the reaction similarities among the GEM models and the pathway models on the basis of topological similarities, metabolite similarities and enzyme commission (EC) numbers. Further analyses can be performed and visualized upon these segments. Grohar is based on open source programming packages and can be obtained freely from https://bitbucket.org/mmoskon/grohar. We demonstrate the application of Grohar on the analysis of cholesterol synthesis pathway in a selected mammalian GEM model.
ViDis: Web-based platform for constructing and sharing visualizations of diagnostic algorithms
Faculty of Computer and Information Science, University of Ljubljana,
The aim of diagnostic algorithms is to aid clinical decisions. The internet provides different sources, in which interested users can browse through different algorithms. These algorithms are, however, dispersed and do not use unified forms of presentation. We introduce Visual Diagnosis (ViDis) tool that provides a web-based platform for construction and sharing of diagnostic as well as other clinical algorithms. ViDis provides an easy-to-use web interface, which makes its functionalities available to a wide scope of users, i.e. doctors and clinicians, researchers and finally to patients. Since it is implemented as a web application it can be accessed through the user's web browser and thus does not require any additional installations. The algorithms are created and displayed in a simplified flow-chart diagram notation. This notation is commonly applied in the field of computer software design and is very intuitive to understand. The platform allows the registered users to contribute their own algorithms. These can be made available to other users, who can browse through all the algorithms in the database using different filters, such as type of the disease or type of the algorithm (e.g., diagnostic algorithm, treatment algorithm or relapse prevention algorithm). Moreover, users can choose to display only curated algorithms, i.e. the algorithms that have been acknowledged by a certified professional. We believe that ViDis provides an excellent platform for knowledge acquisition and dissemination within the medical and research community. ViDis is accessible at http://vidis.fri.uni-lj.si.
A comprehensive understanding of molecular patterns of health and disease is needed to pave the way for personalized medicine and tissue regeneration. One barrier to predictive, personalized medicine is the lack of a comprehensive molecular understanding at the tissue level. As we grasp the astonishing complexity of biological systems (whether single cells or whole organisms), it becomes more and more evident that within this complexity lies the information needed to provide insight in the origin, progression and treatment of various diseases. The best way to capture disease complexity is to chart and connect multilevel molecular information within a tissue using mass spectrometry and data algorithms. It is the realm of big molecular data for disease classification. Charting this territory through the generation of molecular maps from cells and tissue has become reality through the clinical implementation of imaging mass spectrometry complemented with high throughput “omics” approaches. We have demonstrated how new MS based chemical microscopes target biomedical tissue analysis in various diseases as well as other chemically complex surfaces. In concert they elucidate the way in which local environments can influence molecular signaling pathways on various scales. State-of-the-Art molecular imaging with mass spectrometry now enables high resolution tissue screening that provides direct insight into tissue metabolism. Applications have penetrated various research domains from drug metabolism to the visualization of molecular signaling pathways in cancer. This lecture will highlight how mass spectrometry based multimodal molecular imaging can be used to reveal the cellular phenotypes.
Patient's data: a family physician's point of view
The family physician should see a patient as a whole person. This requires broad knowledge, which includes the use of biomedical science but also organisational issues, including management of patient data. This role is becoming increasingly important. Data are increasing and are becoming more and more complex.
The usefulness of data for the patient is a matter of concern. Technology offers possibilities of information which can also cause harm if not interpreted correctly. It is our obligation to inform the patient about his/her risks, which are difficult to interpret.
The ethical dimension is increasingly important: data do not represent only the patient but to an extent also her/his family. Their presentation and storage is ethically sensitive and currently remains an open issue.
A serious issue is the problem of risks that cannot be prevented, where shared knowledge of a risk may gravely influence the patient's behavior and cause anxieties for the patient and the family.
Family doctors need to be equipped with knowledge about the possibilities of new technologies. Together with specialist experts in the field we have to work towards adapting the medical curricula to equip next generations of doctors to face these challenges. This is important especially for family medicine, because family doctors need to know the information that is relevant to them. This calls for a dialogue between different professions in order to come to a common understanding what we can do together.
Computational tools for Nutrition Support in Intensive Care
Aristotle University of Thessaloniki, THESSALONIKI, Greece,
The systems medicine approach in Intensive Care emphasizes on the synergy of computational methods and bedside care, towards improved personalized care to the critically ill. Clinical routine can benefit from data driven predictive models for patient prognosis and treatment.
Adequate nutritional support is among the essential factors for the recovery from critical illness. The daily calories/proteins targets are set by physicians, while scientific evidence is conflicting. In addition, suboptimal and inadequate nutrition is often the case, due to inadequate needs estimation and interrupted nutrition delivery.
Thus, nutrition support is needed in the whole process. To meet this bedside care need, a nutrition tool was developed aiming to calculate optimum nutrition targets/deficits based on current knowledge, monitor nutrition delivery and alert accordingly.
To further leverage smart alerting, a predictive model was developed that alerts for high risk of death for patients that had spent more than 10 days under mechanical ventilation during their total ICU stay. A Neural Network model was trained with features of nutritional data for the first week of hospitalization, plus EHR hospitalization features, to separate patients between those that died vs survived. A 10-fold cross validation took place with 89 patients in balanced classes (45%–55% Death - Survival), with Sensitivity: 74.35% and Specificity 78.26%. In external validation with 14 patients (8 Alive - 6 Deceased) the Sensitivity and Specificity were 75.00% and 83.33%.
This ongoing analysis pinpoints the need for personalised approaches and the value of computational models in bedside nutrition support.
Centre for Systems Medicine,
The key limitation of therapeutics designed to kill cancer cells through the activation of death receptors, such as TRAIL ligands, is susceptibility of the apoptotic pathway to adaptation.
The apoptotic pathway can be initiated from stimulation of numerous different receptor types that differ in terms of cell specific expression level, ligand specific affinities, receptor clustering properties, internalization dynamics and intracellular domain composition. Upon stimulation many of those receptors form vulnerable intracellular protein complexes that activate apoptotic initiator Caspase 8 (Cas8) and/or trigger proliferative signal transduction via the NF-κB pathway. The RIPoptosome, one of those signalling complexes, apart from Cas8 activation also initiates necroptosis by accumulating heterodimers of receptor-interacting proteins (RIP) building filamentous scaffold for activation of the mixed lineage kinase domain-like (MLKL) pseudokinase.
We developed a semi-stochastic model of RIPoptosome formation upon death receptors stimulation that explains how changing level of death receptor-ligand complexes, theirs clustering property and intrinsic molecular fluctuations in RIPoptosome formation drive heterogeneous dynamics of Cas8 activation. The model was implemented as a hybrid of the direct Gillespie stochastic simulation algorithm for slow assembly of RIPoptosome and a deterministic system for the downstream Cas8-Cas6-Cas3 cascade.
Combining our model with experimentally derived set of quantitative protein profiles, literature based catalytic and binding rates acquired for HeLa cells, we have simulated the HeLa cell culture response to death ligand treatment. Based on our modelling results we predicted heterogeneity in response to particular treatment dosages at the single cell level leading to dosage dependent fractional cell death in this population.
Application to severe asthma
The methodological workflow to perform this analysis has been recently published (1).
We selected “omics” data from U-BIOPRED (2) related to blood: whole blood transcriptomics, serum proteomics, urine metabolomics and plasma lipidomics.
Within 372 asthma patients, we identified 8 clusters with minimal deviation from ideal stability (3). The comparison of clinical variables between the clusters highlights expected profiles in asthma but also novel patient subsets.
Preliminary biological interpretation of the clusters shows interesting differences in signalling pathways (e.g. Wnt, MapK, cAMP and Notch) and in immune system related pathways.
Predictive modelling of clusters is being undertaken. The first results show that it is not possible to build models using clinical data only, which is a promising sign that the clusters we found are novel and interesting.
External validation of the models will be the following step; we have secured collaborations with two external cohorts.
Genome Scale Metabolic Models as a new tool in personalized medicine
University of Barcelona, BARCELONA, Spain,
The traditional approach in cancer patient stratification and therapy administration has been based on the examination of specific traits, such as the identification of single mutations or the effect of targeting commonly known genes implicated in carcinogenesis. However, in order to tackle the crucial issues of the efficacy and potential side-effects of widely used treatment combinations there is a need to consider the entirety of events and different levels of regulation implicated in carcinogenesis, together with the distinct genomic fingerprint of each patient.
Nowadays, Genome-Scale Metabolic Models (GSMMs) are widely used to computationally simulate the whole of the biochemical pathways and events that take place in a cell at a genome level. These mathematical models provide an appropriate platform for integrating a variety of different omic data, such as transcriptomics and metabolomics obtained from each patient, and can be used to perform simulations towards designing personalized therapeutic strategies or deciding upon the adequacy of specific treatments at a patient level.
Here we present developments on metabolic modeling tools applied to find new targets to treat cancer patients. Overall, we have improved the GSMM map introducing new constraints determined by metabolomics data. The proposed pipeline is a first step towards the use of GSMMs in clinical decision making, based on patient genome wide transcriptomics and metabolomics data and aims in providing an applicable method for the delivery of personalized cancer therapy and targeted treatment.
Acknowledgements: European Commission (HaemMetabolome, EC-675790), Catalonia Government (2017SGR1033), and Spanish Government-MICINN (SAF2017-89673-R).
Centre for Functional Genomics and Bio-chips, Faculty of Medicine, University of Ljubljana, Slovenia,
Systems approaches contribute to the understanding of complex biological systems such as liver and its surrounding tissues and organs (1). Different adaptations of SteatoNet helped answer different unresolved questions of modern hepatology and provide further insights into metabolic processes, which have not been fully understood.
Hormonal influence on the liver was identified as an extremely important factor in the development of liver related diseases. In this study, SteatoNet was adapted to gender differences in the expression of sex hormones and in dynamics of the growth hormone. The adapted model LiverSex can be used to investigate gender related metabolic differences (2). The LiverSex was validated with experimental data. Sensitivity analysis of the LiverSex showed that communication between the liver and adipose tissue include several critical points in the development of NAFLD in both genders. We proposed VLDL, the storage of triglycerides in lipid droplets, and/or the partition of fatty acids to the ketone bodies as possible mechanisms that protect women from the development of NAFLD.
The customized model SteatoNet with added metabolism of alcohol was named StAlco and was validated with literature data. The StAlco allows the investigation of biochemical consequences of excessive alcohol consumption to liver.
The LiverSex and StAlco present the first computational models for investigating the sex dimorphism and the influence of alcohol on the liver metabolism. They represent excellent starting points for better understanding of the mechanisms between the liver and tissues, in terms of the development and progression of liver-related diseases to specific personal characteristics.
Systems Medicine & Personalised Medicine in Europe -ERACoSysMed and ERAPerMed
ERAPerMed used the IC-PerMed Action Plan and launched its first call aiming at an interdisciplinary approach that combined (pre)clinical research with data and ICT solutions. Patient empowerment, stakeholder participation, communication and dissemination are important activities in ERAPerMed.
ERAPerMed launched the co-funded call early 2018 for a total budget of 27,6M€. Additional calls are scheduled in 2019, 2020 and 2021.
A systems approach to patients with COPD
IDIBAPS, Hospital Clínic of Barcelona, University of Barcelona, BARCELONA, Spain,
We hypothesized high potential for both cost-effective preventive strategies and enhanced clinical management of patients with Chronic Obstructive Pulmonary Disease (COPD) provided that a patient-centred approach is adopted. In this direction, the study aimed to develop subject-specific health risk predictive modelling for early identification of patients showing: i) rapid decay of lung function; ii) high risk of repeated COPD exacerbations; and/or; iii) susceptibility to develop co-morbid conditions and systemic effects of the disease.
We articulated outcomes of research and deployment initiatives carried out in the region of Catalonia (ES) during the period 2012-2017 [Tenyi, A. et al., http://hdl.handle.net/2445/124046], which consisted of iterative processes wherein data from different sources, namely: animal experimentation, human studies, epidemiological research and registry information, were analysed combining several modelling techniques.
Abnormalities in co-regulation of bioenergetics, inflammation and tissue remodelling processes were identified as central players in non-pulmonary manifestations of COPD. The findings showed significant associations with aerobic capacity but not with lung function, with a relevant role for oxidative stress as a key characteristic mechanism. Complementarily, a data-driven analysis concluded in a higher risk of co-morbidities in patients with COPD. Finally, a population-based health risk assessment of 264k COPD cases indicated a significant predictive role of co-morbidities (ROC AUC 0,75-0,85) on relevant clinical events.
Enhanced multilevel predictive modelling that takes into account registry data, underlying disease mechanisms, electronic medical records and informal care information constitute a high priority to pave the way toward personalized medicine for patients with chronic disorders.
Supported by CONNECARE and NEXTCARE.
Physiology & Biochemistry Department, Faculty of Medicine & Surgery, University of Malta, Msida, Malta,
LifeGlimmer GmbH, BERLIN, Germany,
Different data sources ranging from omics data acquired from patient samples as well as animal models to (often handwritten) health records have been collected. Integration and analysis of these data was needed to get a comprehensive understanding of the underlying disease mechanisms.
Hannover Medical School, HANNOVER, Germany,
Digital pathology opens novel opportunities for interactive teaching. Concomitantly, computational pathology raises demand for high-quality image annotation as ground truth for Machine Learning. In context of Systems Medicine research involving spatially resolved “big data from biopsies”, intended to inform mathematical models, we report three years of experience with crowdsourced image annotation as voluntary activity for medical students associated with curricular histopathology courses. The tasks to (1) classify images showing automatically detected tissue structures, and (2) delineate relevant anatomical or pathologically altered tissue compartments help to recapitulate anatomy and address bottlenecks in computational pathology. Own software development and the open-source platform http://www.cytomine.de enabled participation in classroom or remote setting. Technical aspects were iteratively improved. Direct feedback from participants and later on questionnaires on motivation and potential learning benefit allowed optimization. After showing technical feasibility of crowdsourcing by an “educated crowd” of medical students, training material was developed, and an analysis pipeline was created. Voluntary participation rate was 6-20 % of 3rd year students enrolled in the obligatory pathology course. The predominantly positive feedback suggested that crowd sourcing could be accepted as additional interactive training module, provided that it complements well established teaching in pathology. A high proportion of participants indicated as main motivation the opportunity to participate in ongoing research projects, and a common interest in topics related to machine learning, artificial intelligence, and collective knowledge. We conclude that crowdsourcing has a high potential to engage medical students and young physicians in Systems Medicine at an early phase in their career.
The use of unsupervised clustering on omics data already identifies unknown cancer subtypes explaining clinical observations such as survival time. However, the strong bias towards cancer renders many of the subgroup detection methods unfeasible for systems medicine (SM) on other complex diseases. In the latter case, next to high throughput molecular omics data the acquired data includes environmental factors, anthropometric as well as clinical measures. Many of these data sources are of mixed-type, i.e., they contain continuous (e.g., expression data), discrete (e.g., mutation yes/no) and categorical (e.g., ethnicity) values. The integration of the additional detailed patient characteristics demands for the extension of existing or development of new methods that can be used in analyzing complex SM data sets.
In our survey, we describe the current landscape of clustering approaches for biomedical mixed-type data sets. In our contribution, we rely on the methodological reviews from Huang et al. (2017) and Bersanelli et al. (2016) both focusing on multi-omics data. We extend their work in pinpointing the methods that are capable of calculating on mixed-type data sets and introducing twelve additional methods. We also share details on algorithms that were developed in other domains, e.g., social studies. As a result, we highlight theoretical strengths and flaws of the existing methods and show preliminary results of a practical evaluation on SM data assessed within the Systems Medicine Approach for Heart Failure (SMART) project.
Transcriptomic and epigenetic changes in the lung of patients with severe COPD
IDIBAPS, Barcelona, Spain,
Patients with severe COPD have a distinct pulmonary immune infiltrate that should be better characterized integrating differential mRNA expression with epigenetic changes.
The aim of this study was to compare the pulmonary mRNA and miRNA expression profiles of patients with COPD in relation to the severity of airflow limitation. We included in this study 68 former smokers with COPD and 20 never-smokers undergoing thoracic resectional surgery. Total RNA was extracted from whole lung tissue using the miRNeasy kit (Qiagen, Germany). Small RNA sequencing was performed following the Illumina procedure and mRNA expression has been previously assessed with Affymetrix arrays.
We have identified 22 miRNAs differentially expressed between non-smokers and patients with COPD. Additionally, 37 miRNAs were differentially expressed in relation to the severity of the airflow limitation. To integrate miRNA and mRNA data we built two types of networks: 1) miRNA-mRNA spearman negative correlation networks, and 2) networks of putative targets. Several mRNA and miRNA differentially expressed and correlated genes were identified. Currently these results are being validated in a second cohort of patients.
We conclude that there are differentially expressed miRNAs that correlate with mRNA level in lung tissue of severe patients with COPD.
All India Institute of Medical Sciences, New Delhi, India,
The world is poised to enter the post-antibiotic era as per the 2014 WHO report on antimicrobial resistance. Two million new cases of antibiotic-resistant infections are reported every year in the United States alone, with much higher incidence in developing countries. Stewardship programs restrict the indiscriminate use of antibiotics by using summary statistics and antibiograms. However, there is an untapped potential for machine-learning and artificial-intelligence (AI) guided antimicrobial stewardship. We report a Bayesian AI guided stewardship paradigm with that leverages the shared structure of the drug-organism resistance network. Antibiotic susceptibility testing from 2015–2018 were obtained from a pediatric Intensive Care Unit of a tertiary care hospital at Delhi, India. An ensemble-averaged skeleton of the resistance network was learned using Bayesian structure learning and parametrized using Bayesian Information Criterion. Network motifs were exploited to query relevant structures and to make interactive inferences with the potential to influence stewardship decisions. These included actionable insights such as the potential for better utilization of chloramphenicol for less severe infections and cefotaxime being a critical branch-point for taking stewardship decisions. To close the loop of clinical adoption, an interactive dashboard, AMR_Steward, was released as an open-source R package in the public domain (http://doi.org/10.5281/zenodo.1255584
Amsterdam UMC, AMSTERDAM, The Netherlands,
Asthmatics clustered together with a statistically significant subgrouping of pre and post viral challenge states whereas healthy subjects clustered separately with no subgrouping of such states. Asthmatic and healthy responders to virus are similar and cluster together than to non-responders, prior to viral challenge.
Institute of Biomedical Research August Pi Sunyer (IDIBAPS), Barcelona, Spain,
We aimed at applying systems medicine approaches combining integrative omics, imaging and clinical data with computational tools, to developing personalized healthcare for people with MS. We recruited a prospective cohort of 329 MS patients and 90 healthy controls (HC). At baseline, we collected clinical information, and imaging data MRI and OCT. We conducted a multi-omics study including genotyping and calculating the MS genetic burden score (MSGB), cytomics and phosphoproteomics in peripheral blood mononuclear cells (PBMCs) by xMAP assays. We found that the MSGB was significantly higher in patients compared with controls: MSGB: MS 4.23 vs HC 3.2 (p=3.4×10−8), MSGBHLA: MS 1.57 vs HC 0.95 (p=1.6×10−4), MSGBnon-HLA: MS 2.6 vs HC 2.2 (p=6.8×10−5). Regarding differences in immune cells subpopulations we found significant differences in patients compared with controls for regulatory T and B cells. Phosphoproteomics analysis and signaling pathways modeling pointed to an over-activation of the MAPK and NFKB pathways in patients with MS. In summary, our results show significant differences between patients and HC at different scales of biological systems. Such markers will be used for searching prognostic and predictive biomarkers and developing clinical decision support systems for improving disease management.
The ability of cancer cells to evade apoptosis or programmed cell death is a key hallmark of cancer. Understanding this process systematically will improve our understanding of treatment resistance to current radio- and chemo-therapy. Moreover, targeting of apoptosis resistance is a goal of many research teams as this can lead to the development of urgently required novel therapeutics for this disease that target this defective pathway. Novel tools to predict the therapeutic efficacy of current radio- or chemotherapy and to identify new therapeutic approaches such as the inclusion of apoptosis sensitzers would be of significant benefit to patients and their clinical care teams. Over the last decade, several systems biology tools have been developed and validated that model the different biochemical pathways involved in the execution of apoptosis. These systems approaches revealed the ‘all-or-none’ nature of apoptosis execution and analysed the underlying, complex biological networks with key feed back loops and signalling redundancies. Several of these approaches have proven to be of significant prognostic value in the setting of cancer. For the future integration of these approaches into clinical practice, it is important to adapt such systems approaches with routine clinicohistopathological practices. In this presentation, I will review the development of system medicine approaches that model apoptosis signalling, and their potential as prognostic biomarkers and patient stratification tools, with a specific focus on the most aggressive brain cancer, glioblastoma.
Systems Medicine Education at Georgetown University
We are in the midst of “BIG-DATA” era of biology. With the sequencing of the human genome and availability of high power computational methods and various high through-put technologies, biomedical sciences and medicine is undergoing a revolutionary change. The new field of systems medicine is the application of systems biology approaches and tools to biomedical problems. In medicine, complex computational tools will become essential for deriving personalized assessments of disease risk and management including individualized diagnosis, prognosis, and treatment options.
This change, involving the use and analysis of enormous quantities and variety of data, will require a new type of physician and a biomedical scientist with a grasp of modern computational sciences, “-omic” technologies (genomics, proteomics, metabolomics, transciptomics etc.), and a systems approach to medicine. The new tools that clinicians will use continue to arise from the intersection of research across a variety of disciplines and are difficult to capture in traditional curricula. As a leader in medical science and education, Georgetown University Medical Center has taken a leadership role in developing graduate and medical curricula and expanding research in this area.
A dual degree program MD/MS in Systems Medicine was successfully launched in the Fall of 2011. Based on the success of the Dual degree program in the Fall of 2016 we launched, for the first time, a free-standing MS in Systems Medicine and have graduated 12 students. We have developed new courses geared towards training the next-generation of biomedical scientists and physicians. The educational platform will be presented.
IBD-OMICS: A Comprehensive Database for Studying IBD
IBD is a chronic inflammation of all or part of the digestive tract. The two most common and distinct form of IBD include Crohn's disease (CD) and Ulcerative Colitis (UC). Inflammation affects the entire digestive tract in CD while only the large intestine in UC. Both diseases are characterized by an abnormal immune response which has formed the focus of current treatments. While we have come a long way, the etiology and pathogenesis of CD and UC still remain largely unknown. With the explosion of –omic data, systems medicine approaches may lead to identification of biomarkers and newer drug-targets for personalizing the treatment options. We have taken such an approach for understanding CD and UC by extensively mining the literature, and gathering data from several –Omic databases that include Genome Wide Association (GWAS), Universal Protein Resource Knowledgebase (UniprotKB), and microarray expression data. We however realised that a comprehensive collection of –omics data is lacking.
We have created a One-stop-Shop database (IBD-OMICS) that contains gene, protein, snp, biomarker, drugs, expression and relevant articles on IBD, UC and CD. We have also created two tools GDSome and UDsome that is accessible via the database. These tools leverage the Uniprot and GWAS databases and can be used to mine for other diseases that are related to IBD, CD and UC to enable studies geared towards drug-repurposing.
Systems medicine is powerful and is successful in identifying biomarkers and newer targets for IBD. The utility of the database is presented via examples of case studies.
Can big data help to improve the delivery of quality healthcare?
Modern healthcare is facing several challenges, and one of the more pressing challenges is the delivery of quality healthcare. Such quality can be divided along two dimensions. The first is the quality of the technical care itself. Is the healthcare professional sufficiently competent to make accurate diagnoses and then to prescribe efficacious treatment? The second involves an ethical dimension of the healthcare's delivery. Is the healthcare professional providing quality of healthcare that respects the patient's values and preferences? In terms of big data, then, the question emerges, ‘Can big data help to improve the delivery of quality healthcare?’ In this paper, an affirmative answer to that question is advanced and defended. Specifically, big data provides the healthcare professional and patient with more relevant information concerning the patient's illness so that the healthcare professional—along with the patient—can make better informed decisions not only in terms of diagnosis but also with respect to which therapy might be most effective, given the patient's individual illness experience. At first blush, big data can certainly be mobilized to deliver quality healthcare with respect to both its technical and ethical dimensions, but implementation of such data do face several significant challenges, in terms of computational power, data storage, security, and public policy, which appear insuperable but are not. In conclusion, big data do offer the healthcare profession the opportunity to improve its delivery of quality care.
Gene regulatory network modeling of macrophage polarization supports the continuum hypothesis of phenotype differentiation states
CNR Consiglio Nazionale delle Ricerche, ROME, Italy,
Macrophages derived from monocyte precursors undergo specific polarization processes being influenced by the local tissue environment: classically-activated (M1) macrophages, showing a pro-inflammatory activity affecting effector cells in Th1 cellular immune responses; and alternatively-activated (M2) macrophages, with anti-inflammatory functions, involved in immunosuppression and tissue repair. At least three distinctive subsets of M2 macrophages, i.e. M2a, M2b and M2c, are characterized in the literature based on their eliciting molecular signals. The triggering and polarization of macrophages is attained through numerous, interweaved signaling pathways.
To depict the logical relations among the genes involved in macrophage polarization, we utilized a computational modeling methodology, viz. Boolean modeling of gene regulation. We combined experimental data/knowledge from the literature to build a logical gene regulation network model driving macrophage polarization to M1, M2a, M2b and M2c phenotypes. Exploiting the GINsim software we studied the network dynamics under different settings and perturbations to comprehend how they affect cell polarization.
Simulations of the network model, enacting the most significant biological conditions, showed consistency with the experimentally observed behaviour of in vivo macrophages. The model could properly replicate the polarization toward the four main phenotypes as well as to numerous hybrid phenotypes, known to be experimentally associated to physiological and pathological conditions.
We speculate that shifts among different phenotypes in our model mimic the hypothetical continuum of macrophage polarization, with M1 and M2 being the poles of a continuous succession of states. Our simulations also suggest that anti-inflammatory macrophages are more resilient to shift to the pro-inflammatory phenotype.
Immunowell Foundation, UTRECHT, The Netherlands,
What influence the exposome has on a person's health is still a big unknown in biomedical research. The exposome, being the totality of human environmental exposures, including lifestyle factors, covers such a wide variety of aspects that the first question is: where to begin searching for answers?
In this lecture we introduce two methods that provide a way of discovering insights and patterns in how lifestyle and other environmental factors affect health. This in turn can lead to the definition of specific areas that need further research, thus providing a starting point for exploring the exposome. The two methods, Participatory Narrative Inquiry and Large Scale Single Subject Research Design, have not yet or hardly been applied in the field of biomedical research.
They present a structured way of collecting and analysing experiental knowledge of patients. Cutting edge bioinformatics, machine learning and multidimensional modeling using high-performance computing infrastructure is essential to turn collected data into meaningful biological insights for participating patients, medical doctors and scientists. Several examples are given to elucidate the methods and their useful applications in systems medicine.
MacroH2A1 is a histone variant of histone H2A that acts as a barrier to somatic cell reprogramming. We recently demonstrated that loss of macroH2A1 promotes cancer stem cells (CSC) features in hepatocellular carcinoma (HCC), including resistance to chemotherapy and increased glycolysis. How CSCs can influence the neighbouring parental/differentiated cancer cells is unknown.
We obtained RNA-Seq profiles of HCC cells control, KD for macroH2A1 or control exposed to conditioned medium (CM) from macroH2A1 KD cells. Gene ontology (GO) analysis clustered differentially expressed genes between KD, or CM, versus control, as belonging to similar categories such as cancer and gastrointestinal diseases. GO revealed also an enrichment in the categories of cell mediated-immune responses.
MacroH2A1 KD HCC cells could transfer a chemo-resistant phenotype to parental HCC cells in a paracrine manner. To investigate the metabolic changes of CSCs, we used UHPLC to detect metabolites involved in carbon metabolism. 17/64 metabolites were significantly altered in macroH2A1-KD HCC versus control cells. Globally, these changes indicated activation of glycolytic pathways branches that may be linked to altered immune responses detected by RNA-Seq. In fact, a decreased production of IL-8, ENA-78 and MIP-1delta was discovered in the supernatant of macroH2A1 KD cells. CSC cells with low levels of macroH2A1 were very aggressive and failed to activate an adaptive immune response.
In summary the absence of macroH2A1 confers a CSC-phenotype that reprograms proximal HCC cells in a paracrine fashion. Integration of transcriptomic and metabolomics data is a powerful tool for cancer cell phenotyping.
1H-NMR-based prediction of incident type 2 diabetes in the Estonian Biobank
Estonian Genome Center, University of Tartu, Estonia,
We used a machine learning approach to find a set of markers that can best predict the development of type 2 diabetes. For the interpretation of 1H-NMR markers we created a mutual information network integrating metabolomics and proteomics data.
Amsterdam UMC, University of Amsterdam, Department of Respiratory Medicine, Meibergdreef 9, Amsterdam, the Netherlands,
University of Ljubljana, Faculty of Medicine - ELIXIR-SI, LJUBLJANA, Slovenia,
Glioblastoma (GB) is the most aggressive brain malignancy. There is a lack of cell surface markers capable of distinguishing brain tissue from glioblastoma and/or glioblastoma stem cells (GSC), which are responsible for the rapid post-operative tumor reoccurrence.
In order to find new surface GB/GSC markers, we performed meta-analysis of genome-scale mRNA expression data from three data repositories (GEO, ArrayExpress and GLIOMASdb). The search yielded ten appropriate datasets, and three (GSE4290/GDS1962, GSE23806/GDS3885, and GLIOMASdb) were used for selection of new GB/GSC marker candidates, while the other seven (GSE4412/GDS1975, GSE4412/GDS1976, E-GEOD-52009, E-GEOD-68848, E-GEOD-16011, E-GEOD-4536, and E-GEOD-74571) were used for bioinformatic validation.
The selection identified four genes that encode cell surface proteins (CD276, FREM2, SPRY1, and SLC47A1) and the bioinformatic validation confirmed these findings. Results of literature review and bioinformatic validation guided us to consider FREM2 and SPRY1 for experimental validation. This validation revealed that FREM2 expression (but not SPRY1) is generally higher in more aggressive glioblastoma cell lines, such as stem-like GB cell lines and the non-stem cell line U251MG, which represents the aggressive mesenchymal GB type. It is also higher in GB cells lines compared to normal astrocytes. Both FREM2 and SPRY1 are expressed on GB cell surface, while SPRY1 alone was found overexpressed in the cytosol of non-malignant astrocytes.
FREM2 is thus proposed as a novel GB marker and a putative GSC marker, while SPRY1 may be useful as a target for therapeutic antibodies due to its localization on GB cell surface.
Laboratory of Systems and Synthetic Biology, Wageningen University & Research, Wageningen, the Netherlands,
Necrotising fasciitisis (NF) is a severe infection of soft tissue with no clear aethilogy and the mechanisms that drive the infection and the host response to it are not yet clear. In this study we used data from several NF patients admitted to hospitals in Norway, Sweden and Denmark. Dual RNAseq on biopsies of these patients was used to obtain snapshots of the activity of the host, the pathogens and their interaction during the infection. Through multivariate statistics, we correlated the gene expression patterns of host genes with those of the pathogens.
We identified a small number of human genes highly correlated with several different groups of bacterial genes. More than 70% of the bacterial genes in the correlation network are from two important pathogens of NF, Streptococcus pyogenes and Streptococcus dysgalactiae. There are 34 genes from S. dysgalactiae that correlate with one human gene PFKFB3, while 22 genes from S. pyogenes correlate with 7 human genes. This underscores the hypothesis that some (expressed) host genes regulate or influence certain functionalities of the pathogens.
The S. pyogenes genes are involved in several biosynthetic processes that could be of importance in biofilm formation. One of the neighbours of the S. pyogenes genes, PEAK1 gene has been found to play a role in cell adhesion. This cluster could represent the host's defense mechanism. By looking deeper into such relationships revealed by our data-driven approach, the mechanisms behind NF infections can be better understood and potentially lead to new intervention strategies, including faster diagnosis.
Advanced Glycoscience Research Cluster (AGRC), NUI Galway,
In addition to the psychological depressive phenotype, major depressive disorder (MDD) patients are also associated with underlying immune dysregulation which has been shown to correlate with a metabolic syndrome prevalent in depressive patients. Presenting a knowledge base of a multi-tiered biological and psychological entanglement as an ideal clinical condition for a robust integrative analysis of biological pathways underlying the dysregulated neural connectivity and systemic inflammatory response, which may aid in devising effective strategies to treat and alleviate associated comorbidities.
We explored an integrative systems biology methodology by combining genomics data on MDD patients from public databases (n=26), meta-analysis of biomarker data from published articles (n=35, clinical studies) and results from a major genome-wide association meta-analysis (n=44 variants in 130,664 MDD cases). By leveraging online software tools (GenePattern and GALAXY), we identified the differentially expressed genes (fold value ≥2 or ≤−2 fold, p≤0.05, FDR=0 – 0.1) from both blood and brain samples, comparing its expression pattern with the anti-depressant drug treated conditions (3 non-human and 2 human datasets of MDD patients).
Detailed gene set enrichment analysis and complex protein-protein, gene regulatory and biochemical interaction pathway analysis were performed using online software tools (NIH LINCS program and Consensus PathDB) and our proprietary GlycoGAIT database, to identify the functional significance of these markers. This integrative analysis method provides insights into the molecular mechanisms along with key glycosylation dysregulation underlying altered neutrophil-platelet activation and dysregulated neuronal survival maintenance and synaptic functioning which are supported across literature to be dysregulated in MDD patients.
Novo Nordisk Foundation Center for Protein Research, University of Copenhagen, COPENHAGEN, Denmark,
Breast cancer is a heterogeneous disease, usually divided into molecular subtypes. Some patients within subtypes do not respond to treatment, showing the need for better stratification. We analysed temporal data (January 2006–June 2016) consisting of laboratory tests for thousands of breast cancer patients that mainly consisted of: lab tests, clinical notes, coding of diseases and drugs for around 3 million people. In this cohort around 30,000 were diagnosed with a breast cancer diagnosis. We took inspiration from network biology and single-cell sequencing approaches, to build a pipeline for clustering patients using lab tests. An important step of the pipeline was the use of the PhenoGraph method. The key step in the PhenoGraph method is to convert the lab data to a graph that represents the phenotypic similarities between patients. Then, it calculates the Jaccard coefficient between nearest-neighbor sets, builds an undirected graph from the weighted links and identifies communities using the Louvain method on the graph. The analysis of multiple lab tests and time points, showed that traditional lab test data are able to discriminate several clusters of patients according to different disease progression and other specific temporal trends. The approach was able to identify putative and novel subgroups that could increase our knowledge about breast cancer subtypes. The introduction of lab test results for a new patient into the method, could contribute to the early prognosis and identify the subgroup the patient belongs to earlier and could help in the decision-making process leading to the right treatment.
Institute of Clinical Molecular Biology, UKSH, Kiel, KIEL, Germany,
Vedolizumab is a monoclonal antibody used for the treatment of inflammatory bowel disease (IBD). Vedolizumab targets the integrin heterodimer α4β7, which facilitates leukocyte migration to the intestinal mucosa. Inhibition of leukocyte trafficking to the inflamed mucosa is suggested to be the reason for the efficacy of vedolizumab, however clinical studies supporting this conclusion are lacking. Here, we aim to test the effects of vedolizumab on mucosal and systemic immunity, and to understand the molecular changes linked with the efficacy of vedolizumab.
Peripheral blood and mucosal biopsies were collected from IBD patients, who received vedolizumab therapy, before and 2, 6 and 14 weeks after therapy, and comprehensive analyses of immune cell populations, T-cell receptor repertoire and RNA-sequencing data were performed. We found that vedolizumab only has minor effects on intestinal T-cell abundance and mucosal T-cell receptor repertoire in the patients. Moreover, cell deconvolution analysis based on transcriptomic data showed alterations in the relative abundances of macrophages but not T-cells, B-cells or plasma cells before and after treatment. Differential gene expression analysis identified a higher number of downregulated genes in patients who attained remission, indicating an inhibitory effect on immune pathways. Gene Ontology analysis of these genes revealed highly significant enrichment of terms “innate immune response” and “inflammatory response”. These findings suggest that modulation of innate immunity contributes to the therapeutic efficacy of vedolizumab in IBD. We further aim to identify unique signatures of remission in response to vedolizumab that can be used to predict the therapy outcome.
University Hospital Groβhadern, MÜNCHEN, Germany,
St'Anne University Hospital Brno - International Clinical Research Center, BRNO, Czech Republic,
Nowadays, “omic” studies and parabiotic screens have identified molecules that can reverse or slow down aging and Growth Differentiation Factor 11 (GDF11) has attracted considerable attention. GDF11 belongs to the TGFβ superfamily and plays a pleiotropic role throughout the mammalian development. The role of GDF11 in aging remains controversial. Mice studies have shown that systemic restoration of GDF11 reverses age-related phenotypes in the heart, skeletal muscle, pancreas and the brain. However, other evidence suggests that its supraphysiological concentrations induce muscle waste and death. The role of GDF11 in hepatic lipid metabolism is unknown.
In this study, we investigated the effect of GDF11 on free fatty acids (FFA)-induced changes in lipid accumulation and composition in hepatocytes. Our confocal microscopic experiments demonstrated smaller lipid droplets in HepG2 cells treated with GDF11. RNA-Seq analysis showed that GDF11 is able to upregulate its own transcript in positive feedback loop. Endogenous metabolic profiles were determined by ultra-high performance liquid chromatography coupled to mass spectrometry (UHPLC-MS), optimized for profiling glycerolipids, glycerophospholipids, sphingolipids and cholesteryl esters.
197 lipid metabolic features were detected in the analyzed samples and the significant differences were found between the treatment groups. The metabolites responsible for the observed differences were mainly sphingomyelins, diacylglycerols, triacylglycerols and glycerophospholipids. GDF11 treatment led to the significant decrease of triacylglycerols and ceramides and increase of phosphatidylcholines when compared to control and FFA-treated cells.
In conclusion, our data suggest that GDF11 has effect on lipid metabolism and formation of lipid droplets, which could be harnessed for clinical applications.
Metabolic processes are regulated differently in the 24h period of the day-night cycle. One of these processes is cholesterol synthesis, which is highly regulated by circadian rhythm. Besides that, we propose that cholesterol or its synthesis intermediates feedback to the circadian clock genes and regulate their expression. To test this hypothesis we applied the mouse embryonic fibroblasts (MEF) with inactivated Cyp51 gene that encodes an enzyme in the late part of the cholesterol synthesis. MEFCyp51-/- cells are unable to synthesize cholesterol and the concentration of cholesterol intermediates is altered. We first characterized the immortalized MEF cells on DNA, mRNA and protein level to confirm the knockout. RNA was isolated from cells at different time points during the 24-h period and the expression of clock genes was quantified. To experimental data, we mathematically aligned the cosine fitting curve, to obtain the graph of circadian expression of clock genes and predict expression at the time points not obtained experimentally. MEFCyp51-/- cells have an elevated amplitude of the core clock genes (Per2, Bmal). We suggest that transcription factor RORC is responsible for the altered circadian gene expression when cholesterol synthesis is changed. RORC on one hand regulates the expression of the core clock genes, while on the other hand it is activated by sterol intermediates of cholesterol synthesis, whose concentrations are altered in MEFCyp51-/-. We propose that cholesterol homeostasis can fine-tune the circadian clock through the RORC signaling pathway which can further influence other metabolic pathways that are regulated by the clock.
Information System for the Targeted Microbiome Correction by Nutrition and Pharmabiotics
Human microbiome determines our personal health, thus prognostic correction of its composition to expected functional activity by pharmabiotics and/or nutrition is a promising approach for prevention and treatment of noncommunicable diseases.
Because of the uniqueness of microbiome of a particular person and its flexibility, the efficacy of such “medications” implementation depends on the level of personification in the selection of diet/pharmabiotics components.
We have developed information system (IS) and began initial [pilot] testing of its beta-version for the adjustment of microbiome ratio correction in correlation with other biomarkers through the selection of ingredients rich in biologically active molecules (BAM) and for the design of nosology – and patient- specific pharmabiotics.
The IS operates with big data, in particular: i) patients anamnesis, medical examination results, ii) biomarkers, iii) microbiome representatives ratio for various diseases, iv) BAM of edible plants, v) influence of BAM, food ingredients, microorganisms on microbiome representatives ratio, vi) food composition databases, etc.
Appropriate databases (DBs) have been created and a number of tools have been developed. Part of the DBs are arranged as relational DBs tables, others are organized through external DBs access. The IS has a friendly graphical interface for users allowing quick and easy individual data entry to perform personalized nutrition choices and recommend pharmabiotics.
Innovative algorithm has been developed for the selection of individually required contents from extensive data. Further development of the IS tool is oriented to precise diagnostics at earlier stages of diseases and improvement of individual recommendations by machine learning techniques.
Hepatocelullar carcinoma: the consequence of the disrupted cholesterol synthesis in transgenic mice
Hepatocellular carcinoma (HCC) is considered as a dynamic disorder of genetic, molecular and cellular networks/pathways that are in close connection with the environment. Hepatocarcinogenesis is linked to increasing frequency of steatohepatitis (NASH) that is difficult to diagnose and treat due to lack of specific drugs. To come closer towards understading the molecular players in the progression of NASH towards HCC we applied powerful post-genomic technologies with bioinformatics support and a mouse model - the hepatocyte specific knockout of lanosterol 14α-demethylase (CYP51) (HCyp51-/-) that has a phenotype similar to NASH, with inflammation, fibrosis finally resulting in HCC.
Histopathology and transcriptome analyses were performed on the liver and biochemical parameters measured in plasma. Gene expression analysis was done in R using Bioconductor packages limma and PGSEA. Gene sets for enrichment analysis were taken from KEGG and TRANSFAC databases and validated by qPCR and immunohistochemistry.
First HCC cases were observed at 12-month in KO mice. At 24-months the prevalence of tumors was higer in females compared to the males (88.8%/50%). Hepatocellular damage was confirmed with increased AST and ALT and patohistology. Liver injury in HCyp51-/- mice promoted pathways in cancer and extracellular matrix interaction. Higher susceptibility for tumor developement in older females could be reasoned by sex-specific activation of β-catenin/Wnt and Tgf-β1 signaling. Metabolic pathways, including the RorC signaling, were dampened. Our data indicate that the transcription factors Sox9, Sp1, Ap-1 and RorC are crucial for the progression of NASH towards HCC. Additionally, we have uncovered novel cholesterol-linked sex-specific pathways of hepatocarcinogenesis.
Lifelong - e-learning platform for Systems Medicine
University of Ljubljana, Faculty of Medicine,
ELIXIR is an European infrastructure for life science information. ELIXIR coordinates, integrates, and sustains bioinformatics resources across its members and enables users in academia and industry to access services vital for their research. ELIXIR-SI is Slovenian national node that provides training and bioinformatic tools and services for better use in medicine, biology and other life sciences. A special goal of ELIXIR-SI is to establish a regional training centre. Our approach to training is based on modern online e-learning approach which is more effective compared to the face-to-face learning. To this end we use open source Learning Management System Moodle on which we are developing our own learning platforms named ELIXIR-SI eLearning Platform (EeLP) for ELIXIR and Lifelong for other areas including Systems Medicine. Idea is to cover needs of students and teachers on one point for synchronous as well as for asynchronous courses. The Lifelong platform functionalities covers advanced user (student and teacher) management, AAI authentication using eduGAIN and ELIXIR AAI, communication tools like forums and instant messaging, advanced and standard teaching materials like lessons, and assessments and embedded connection with the tools and services necessary for education and teaching like embedded Web terminal and Galaxy. Within Lifelong we prepared a course about using Galaxy for interactive construction and execution of workflows in the field of Systems Medicine with simple hands-on exercises for the students that was tested during CASyM workshops. During presentation we will outline our current experiences and highlight further development.
Cooking is the core of our cultural identity apart from being the source of nutrition and health. Food interacts with the body in a complex manner leading to health consequences. Data-driven investigations of gastronomic questions from the Complex Systems Laboratory, IIIT-Delhi, have opened up exciting directions for the study of patterns in traditional recipes, their flavor composition, and health associations, setting the foundation of an all-new field—Computational Gastronomy. This emerging interdisciplinary science involves the collection, curation, and analysis of culinary data using methods in machine learning, natural language processing, pattern mining, and chemoinformatics among others. Along with complementary experimental studies, such explorations have the potential to transform the landscape of food by effectively leveraging it for better health and nutrition through an array of culinary applications. This talk will provide an overview of computational gastronomy research involving analysis of food pairing, culinary fingerprints, the chemical and sensory space of flavor compounds (FlavorDB), health impacts of food ingredients (SpiceRx, DietRx), and taste prediction (BitterSweet), other than articulating the future prospects. Beyond the reductionist approach with which food has been hitherto studied for its chemical constitution, interaction with sensory mechanisms, and effect on health, I argue for the relevance of data-centric, systems approach to pave the way for better health through dietary interventions. As the world deals with the problems of food security and diet-linked lifestyle disorders, Computational Gastronomy provides sustainable solutions via data-driven food innovations for personalized nutrition and better health within the cultural context.
iSymbiol: Information system for storage and management of biological samples
Centre for Functional Genomics and Bio-Chips, LJUBLJANA, Slovenia,
Advancements in research methods and technology mean that more data and samples are collected from experiments. Amounts of data and physical biological samples translate to difficulties with their storage, tracking, management and analysis.
We describe iSymbiol, an information system that addresses the aforementioned issues. We use a centralized database environment to store all vital sample information, and which presents a perfect substitution for paper records and non-standardized electronic spreadsheets. Physical identification and tracking of samples is realized with labels containing optical code, which provide unique identifiers of samples and contain necessary sample information. Management of sample data is implemented through a web-based graphical interface which also serves as a way to input and display of data. Input of data can be done manually or imported from existing files. Information system is able to export stored information in easily readable and structured files, which can be further analyzed with number of commercially or freely available tools.
Information systems' ease of use, low maintenance, development with open-source tools and platforms provides an alternative to commercial software solutions and makes it more accessible to scientific and research communities. It was designed as a modular and flexible system that could be easily customized to the users' demands. The system has already been applied in several projects, namely in the diagnostics of Alzheimer's disease and in the unravelling of genetic predispositions of Familial erythrocytosis, hepatocellular carcinoma and suicide tendencies.
One of the cornerstones of precision medicine is comprehensive knowledge retrieval about the targeted individuals. Molecular identification and characterization of patients is thus a high priority. Extracting insights and information from multiple heterogeneous and interdependent data, requires new systems analytics, moving beyond classical algorithmic or mechanical processes. In this presentation, we show how the combination of different viewpoints are needed to translate systems knowledge into day-to-day precision medicine. We give examples at the level of study design development and at the level of advanced analytics exploitation.
Precision-Panc: From molecular profiling to precision medicine in pancreatic cancer
A major challenge inherent to lower incidence cancer types (ranking 5th and lower) is that a network approach is required to make significant advances through exposure to greater capacity and patient numbers. This is particularly the case for pancreatic cancer which although being the tenth in incidence, is the third, and soon to be the second leading cause of cancer death. Our increasing appreciation for the molecular diversity of cancer further exemplifies the need for a networked platform approach. To address this, we established Precision-Panc in the UK. Precision-Panc is a synergistic and dynamic therapeutic development platform aligning ‘discovery’, ‘pre-clinical’ and ‘clinical’ therapeutic development to form a continuous loop of discovery, learning, refinement, and implementation through efficient forward and backward translation. In this presentation, the initial experience of Precision-Panc will be presented, including the translational of molecular profiling and preclinical evidence into the initial suites of clinical trials. We will also discuss the trials in design.
Impact of microbiota on immune responses to vaccination in healthy adults
Emerging evidence demonstrates a major role for the microbiota in shaping physiology. Several elegant studies have demonstrated potent effects of the gut microbiome on modulating immunity in mice, and described striking correlations between particular species of bacteria and various inflammatory diseases in humans. However evidence for a causal effect of the microbiome in shaping immunity in humans is sparse. This issue is of particular relevance with vaccination as variations in the efficacy of vaccines in geographically distinct populations have been thought to be dependent on differences in the gut microbiome. Thus in order to determine the impact of the gut microbiome on vaccine immunity, we performed clinical trials in which broad spectrum antibiotics were administered to healthy adults, prior to vaccination with the seasonal influenza vaccine. We performed extensive profiling of innate and adaptive immune responses in antibiotics-treated and control subjects vaccinated with the inactivated influenza vaccine over two separate seasons. The results of this study will be presented.
A New Integrated Platform for Training in Data Science for Medicine
Big data and a surge in quantitative methodologies produce digital data in unprecedented amounts. Making sense of available data (not just ‘having data’) becomes the main challenge for clinical and biomedical researchers. Training in computational skills is therefore quintessential for successful academic and clinical careers. I present an eTraining platform for Medical Data Science based on most recent web technologies. It allows for standalone on-line training courses as well as for face-to-face workshop support. We cover a range of topics from simple data handling to Network Medicine and Machine Learning. I will report on our experiences with training of clinical researchers from the University College London and on plans to expand the UK.
Signals as vehicles in the assessment of metal-induced cellular pathophysiologies. Theoretical and experimental correlations linked to holistic processing and diagnostic tools in disease
It is well-known that both central (CNS) and peripheral nervous systems (PNS) are exposed to numerous stress-factors. Among them, metal ions, essential or non-essential, may exhibit toxic effects, when their homeostasis is disturbed. Copper (Cu) is such an essential trace element, present throughout the CNS (e.g. hippocampus), thereby being implicated (in)directly in the pathogenesis of several neurological disorders (Alzheimer's disease, etc.). To examine the neurotoxic effect(s) of Cu(II), a rat-isolated sciatic nerve preparation was employed in a systematic ex vivo study. The work targeted potential correlations between viability and Compound Action Potential (CAP) generation of excitable nerve fibers through monitoring CAP, using standard electrophysiological methods. Four intervention groups were considered: control, low (1–75μM), medium (125–250μM) and high-exposure (500μM). Mean values of measurements in each group were calculated hourly, allowing exploration of differences among groups over time. A decrease in a) peak value, distinct in medium/high, b) integral, c) slope, compared to control and drift in latency were observed (even within 4h). The employed water-soluble Cu(II)-species exhibited significant concentration-dependent (1–500μM) reduction in CAP amplitude. Data were analyzed in R. After preprocessing (normalization, noise filtration, artifact rejection), several parameters were calculated (Peak, Integral-surface, Duration, Max-Slope to peak, and Latency to peak) for all consecutive CAPs acquired, up to 12h post-exposure. The discoveries in this study shed light onto key aspects of clinically important conditions, provide a holistic view of cellular pathophysiology, and proffer merit toward further research and employment of safe metallodrugs [e.g. Cu(II)] against aberrational processes in human physiology.
