Filippo Castiglione

OpenMP parallelization of agent-based models

Agent-based models, an emerging paradigm of simulation of complex systems, appear very suitable to parallel processing. However, during the parallelization of a simulator of financial markets, we found that some features of these codes highlight non-trivial issues of the present hardware/software platforms for parallel processing. Here we present the results of a series of tests, on different platforms, of simplified codes that reproduce such problems and can be used as a starting point in the search of a possible solution.

A Model of Ischemia-Induced Neuroblast Activation in the Adult Subventricular Zone

We have developed a rat brain organotypic culture model, in which tissue slices contain cortex-subventricular zone-striatum regions, to model neuroblast activity in response to in vitro ischemia. Neuroblast activation has been described in terms of two main parameters, proliferation and migration from the subventricular zone into the injured cortex. We observed distinct phases of neuroblast activation as is known to occur after in vivo ischemia. Thus, immediately after oxygen/glucose deprivation (6–24 hours), neuroblasts reduce their proliferative and migratory activity, whereas, at longer time points after the insult (2 to 5 days), they start to proliferate and migrate into the damaged cortex. Antagonism of ionotropic receptors for extracellular ATP during and after the insult unmasks an early activation of neuroblasts in the subventricular zone, which responded with a rapid and intense migration of neuroblasts into the damaged cortex (within 24 hours). The process is further enhanced by elevating the production of the chemoattractant SDf-1α and may also be boosted by blocking the activation of microglia. This organotypic model which we have developed is an excellent in vitro system to study neurogenesis after ischemia and other neurodegenerative diseases. Its application has revealed a SOS response to oxygen/glucose deprivation, which is inhibited by unfavorable conditions due to the ischemic environment. Finally, experimental quantifications have allowed us to elaborate a mathematical model to describe neuroblast activation and to develop a computer simulation which should have promising applications for the screening of drug candidates for novel therapies of ischemia-related pathologies.

Characterisation of the Idiotypic Immune Network Through Persistent Entropy

In the present work we intend to investigate how to detect the behaviour of the immune system reaction to an external stimulus in terms of phase transitions. The immune model considered follows Jerne’s idiotypic network theory. We considered two graph complexity measures—the connectivity entropy and the approximate von Neumann entropy—and one entropy for topological spaces, the so-called persistent entropy. The simplicial complex is obtained enriching the graph structure of the weighted idiotypic network, and it is formally analyzed by persistent homology and persistent entropy. We obtained numerical evidences that approximate von Neumann entropy and persistent entropy detect the activation of the immune system. In addition, persistent entropy allows also to identify the antibodies involved in the immune memory.

The onset of type 2 diabetes: Proposal for a multi-scale model

Background Type 2 diabetes mellitus (T2D) is a common age-related disease, and is a major health concern, particularly in developed countries where the population is aging, including Europe. The multi-scale immune system simulator for the onset of type 2 diabetes (MISSION-T2D) is a European Union-funded project that aims to develop and validate an integrated, multilevel, and patient-specific model, incorporating genetic, metabolic, and nutritional data for the simulation and prediction of metabolic and inflammatory processes in the onset and progression of T2D. The project will ultimately provide a tool for diagnosis and clinical decision making that can estimate the risk of developing T2D and predict its progression in response to possible therapies. Recent data showed that T2D and its complications, specifically in the heart, kidney, retina, and feet, should be considered a systemic disease that is sustained by a pervasive, metabolically-driven state of inflammation. Accordingly, there is an urgent need (1) to understand the complex mechanisms underpinning the onset of this disease, and (2) to identify early patient-specific diagnostic parameters and related inflammatory indicators. Objective We aim to accomplish this mission by setting up a multi-scale model to study the systemic interactions of the biological mechanisms involved in response to a variety of nutritional and metabolic stimuli and stressors. Methods Specifically, we will be studying the biological mechanisms of immunological/inflammatory processes, energy intake/expenditure ratio, and cell cycle rate. The overall architecture of the model will exploit an already established immune system simulator as well as several discrete and continuous mathematical methods for modeling of the processes critically involved in the onset and progression of T2D. We aim to validate the predictions of our models using actual biological and clinical data. Results This study was initiated in March 2013 and is expected to be completed by February 2016. Conclusions MISSION-T2D aims to pave the way for translating validated multilevel immune-metabolic models into the clinical setting of T2D. This approach will eventually generate predictive biomarkers for this disease from the integration of clinical data with metabolic, nutritional, immune/inflammatory, genetic, and gut microbiota profiles. Eventually, it should prove possible to translate these into cost-effective and mobile-based diagnostic tools.

The role of computational models of the immune system in designing vaccination strategies.

Mathematical and computational models are designed to improve our understanding of biological phenomena, to confirm/reject hypotheses, and to find points of intervention by altering the behavior of the studied systems. Here we describe the role of mathematical/computational models of the immune system. In particular, we analyze some examples of how mathematical modeling can contribute to finding optimal vaccination strategies. Indeed, computational modeling offers an intriguing opportunity from the theoretical point of view, and it will be of interest for clinically oriented investigators who wish to find optimal therapeutic strategies and for pharmaceutical industries that want to produce effective and successful drugs.

The clinical significance of HCV core antigen detection during Telaprevir/Peg-Interferon/Ribavirin therapy in patients with HCV 1 genotype infection.

BACKGROUNDnDirect-acting antiviral drugs (DAA) regimen improve the SVR rate. However, adverse effects often lead to therapy interruption. This underlines the importance to find some predictive parameters of response in order to consider the possibility of a shorter time of antiviral treatment in the appearance of adverse effects without affecting the success of the therapy.nnnOBJECTIVESnWe aimed to examine the HCVAg kinetics in the early phase of treatment and its predictive value of SVR in patients undergoing TPV/Peg-IFN/RBV treatment.nnnSTUDY DESIGNnTwenty-three patients infected by HCV genotype 1 (1a n=11; 1b n=12) were included in this prospective study.nnnRESULTSnAt baseline the median Log of HCVAg concentration in RVR and EVR patients were 3.15 fmol/L and 3.45 fmol/L, respectively with no significant differences. The baseline median HCV-RNA to HCVAg ratio was 233.77, this ratio was significantly lower when measured on day 1 (27.52) and on day 6 (24.84) (p<0.001). The two-tailed Fishers exact test indicated that the SVR response is statistically significantly different in patients with detected HCVAg at week1 compared to patients with no detectable HCVAg (p=0.05). The sensitivity, specificity, and negative and positive predictive values (NPV, PPV) were 53.8, 87.5, 53.8 and 87.5%, respectively. The area under the ROC curve was 0.71 at day T6, the best cut-off of 3 fmol/L when evaluated with the HCVAg plasma concentration at day T6.nnnCONCLUSIONnUndetectable HCVAg in the early phase of TPV/Peg-IFN/RBV treatment could represent an important parameter for predicting SVR.

Plasma HIV RNA decline and emergence of drug resistance mutations among patients with multiple virologic failures receiving resistance testing-guided HAART.

Early recognition of virologic failure in patients with extensive drug resistance receiving salvage-HAART is essential to avoid exposure to subinhibitory regimens. We studied plasma viral load (PVL) decline and rates of drug-resistance mutation (DRM) accumulation in such patients. A prospective, 48 week study of 38 heavily pretreated patients receiving genotypic resistance testing (GRT)-guided HAART was conducted. The rate of PVL decline was studied by weekly PVL determinations. To assess DRM accumulation, serial GRTs were performed in all nonresponders (never reaching PVL <50 or two PVLs >50 copies/ml after suppression). Over 48 weeks, 10 patients (26%) were nonresponders. Receiving less then two fully active drugs and having an elevated number of PI and NRTI mutations at baseline were strongly associated with virologic failure. There was no evidence of a difference in the change from baseline PVL to week 1 and 2 between responders and nonresponders. By contrast, PVL reductions from week 2 to week 3 and thereafter were significantly greater for responders (p < 0.01). Among nonresponders, the incidence rates per patient-month (95% CI) of emergent DRM were 0.67 (0.13-1.20), 0.40 (0.00-0.74), and 0.37 (0.00-0.75) at weeks 4, 8, and 24, respectively. Having limited baseline resistance, receiving at least two fully active drugs, and showing constant PVL reductions from week 2 to week 3 and thereafter were predictive of virologic response. In contrast, early changes in PVL levels were not. Virologic failure was associated with detection of emergent DRMs. Virologic rebound in patients on salvage-HAART should be addressed aggressively.

Multi-scale Simulation of T Helper Lymphocyte Differentiation

The complex differentiation process of the CD4+ T helper lymphocytes shapes the form and the range of the immune response to different antigenic challenges. Few mathematical and computational models have addressed this key phenomenon. We here present a multiscale approach in which two different levels of description, i.e. a gene regulatory network model and an agent-based simulator for cell population dynamics, are integrated into a single immune system model. We illustrate how such model integration allows bridging a gap between gene level information and cell level population, and how the model is able to describe a coherent immunological behaviour when challenged with different stimuli.

Network Inference and Reconstruction in Bioinformatics

Abstract Systems biology focuses on the integration of experimental, mathematical and computational techniques to develop systemic views and predictive models of biological systems. In this perspective, the concept of network has been a powerful tool for the representation and the analysis of complex systems: during the last two decades, the so-called network biology approach has been fruitfully applied in many different biological areas, from gene regulation, to protein-protein interactions, to neural signals. Here, making no claim to completeness, we briefly account for the processes of reconstructing several among the most significant types of biological networks in molecular biology and neuroscience, as well as for some of the most promising methodologies applied in the recent field of network medicine.

Physio-Environmental Sensing and Live Modeling

In daily life, humans are constantly interacting with their environment. Evidence is emerging that this interaction is a very important modulator of health and well-being, even more so in our rapidly ageing society. Information and communication technology lies at the heart of the human health care revolution. It cannot remain acceptable to use out of date data analysis and predictive algorithms when superior alternatives exist. Communication network speed, high penetration of home broadband, availability of various mobile network options, together with the available detailed biological data for individuals, are producing promising advances in computerized systems that will turn information on human-environment interactions into actual knowledge with the potential to help make medical and lifestyle decisions. We introduced and discussed a key scenario in which hardware and software technologies capable of simultaneously sensing physiological and environmental signals process health care data in real-time to issue alarms, warnings, or simple recommendations to the patient or carers.