Error message

Monday, 01 July 2019
Time Speaker Title Resources
09:00 to 10:00 -- Registration & Welcome
10:00 to 10:50 Jonathan Carruthers Stochastic dynamics of Francisella tularensis infection and replication

Francisella tularensis is a highly infectious bacterium capable of causing a debilitating disease with as few as 10 organisms, and is currently classified as a category A biothreat agent. To study disease progression, we begin with a stochastic model of a population of infectious agents inside a single host cell. Different approaches are considered to determine the time until rupture of an infected macrophage and the number of bacteria released when the cell bursts. From a single-cell model we are able to approximate well the dynamics within the lung of an infected mice, comparing our results with those of agent-based computation. Further comparisons with experimental measurements, carried out after murine aerosol infection with the virulent SCHU S4 strain, enable us to infer model parameters using Approximate Bayesian Computation (ABC), obtaining a bacterial growth rate that is consistent previous beliefs that the time between rounds of infection is less than 6 hours in vivo

10:50 to 11:20 -- Coffee break
11:20 to 12:10 Sara Jabbari Novel strategies to tackle bacterial infections: targeting adhesion and persistence

The ability of bacteria to become resistant to previously successful antibiotic treatments is an urgent and increasing worldwide problem. Solutions can be sought via a number of methods including, for example, identifying novel antibiotics, re-engineering existing antibiotics or developing alternative treatment methods. The nonlinear interactions involved in infection and treatment render it difficult to predict the success of any of these methods without the use of computational tools in addition to more traditional experimental work. We use mathematical modelling to aid in the development of anti-virulence treatments which, unlike conventional antibiotics that directly target a bacterium’s survival, may instead attenuate bacteria and prevent them from being able to cause infection. Many of these approaches, however, are only partially successful when tested in infection models. We present two such potential treatments in relation to the multi-drug resistant bacterium Pseudomonas aeruginosa: targeting host-cell adhesion and cell-morphology transitions that facilitate persister-like behaviour. Using mathematical modelling we consider ways to optimise the efficacy of such treatments.

12:10 to 13:00 Rebecca Chisholm Understanding the relationship between the epidemiology of, and immune response to, Group A Streptococcus infection

Group A Streptococcus (GAS) is a ubiquitous bacterial pathogen that exists in many distinct strains and is highly prevalent in Indigenous and other disadvantaged populations. It is responsible for a range of diseases – from superficial infections of the skin and throat, to life-threatening invasive infections, and post-infection sequelae – and the relative prevalence of each of these diseases differ across populations. Vaccines against GAS are under development, but their effective use will require better understanding of how immunity develops following infection. In this presentation, I will discuss how we are using mathematical modelling in conjunction with local and global epidemiological data to gain insights into the within-host infection and immunity dynamics of GAS.

13:00 to 14:15 -- Lunch
14:15 to 15:05 Sitabhra Sinha Games, Networks and Public Health: The source of disease prevalence information impacts effectiveness of vaccination programs

The recent reports of dramatic increase in measles cases worldwide - particularly in the world’s most advanced economy, USA - is puzzling because an effective vaccine for the disease has existed for a long time. The answer seems to be that the very success of vaccination in reducing the burden of many diseases has paradoxically led to a decline in the number of people willing to get vaccinated. Thus, people think that they are now safe from an epidemic and can avoid getting vaccinated themselves (or their children). Can apparently rational decisions by individuals, viz. not to vaccinate, potentially lead to a catastrophic consequences for society, viz., opening it upto deadly epidemics ? In this talk we show how the decision of individuals to vaccinate themselves is influenced by their risk perception regarding contracting the disease, which in turn in based on information they have about epidemic incidence and the fraction of neighbors on their social network who are protected through vaccination. Using a game theoretic framework we show that optimal public health outcomes arise when individuals use information about disease prevalence in the local neighborhood of their social network, in contrast to relying on global prevalence information obtained from mass media. Our results strongly suggests the need for a transparent system of disseminating detailed incidence information about an ongoing epidemic to the public, such that individuals can make informed vaccination decisions based on real-time data for their neighborhood. This is a collaboration with Dr Anupama Sharma (IMSc Chennai), Dr Shakti N Menon (IMSc Chennai) and Dr V Sasidevan (IMSc Chennai & CUSAT Cochin).

Reference:
Sharma A, Menon SN, Sasidevan V, Sinha S (2019) Epidemic prevalence information on social networks can mediate emergent collective outcomes in voluntary vaccine schemes. PLoS Comput Biol 15(5): e1006977

15:05 to 15:55 David R Sinclair Expected size of measles outbreaks caused by vaccination exemptions for school children

Measles was eliminated, meaning an end to continuous transmission, from the US in 2000. However, outbreaks continue to occur due to introductions from international travel. Measles immunization is mandatory for all school children, unless they have an exemption. Exemptions for personal and religious reasons have grown in the past decade. Geographic clustering of unvaccinated children in certain schools facilitates the spread of measles introductions, but the potential size, and thus risk, of outbreaks is unclear.
I will discuss our efforts to estimate potential outbreak sizes using an agent-based model, populated with a synthetic representation of the US state of Texas. Real schools are represented in the simulations and the current vaccination rate of each real school was applied to its in silico equivalent.

15:55 to 16:25 -- Coffee break
16:25 to 17:15 Antonio Gómez Corral A comparative analysis between two time-discretized versions of the SIS epidemic model

The talk is concerned with \emph{time-discretized} versions of the SIS (susceptible $\to$ infected $\to$ susceptible) epidemic model, which are derived by inspecting the number of infective hosts at a finite sequence of times $\tau_0=0<\tau_1<...<\tau_{m-1}<\tau_m=t'$, for a predetermined time $t'>0$ and an integer $m\in\mathbb{N}$. The aim is to construct a suitable criterion allowing us to summarize appropriately the dynamics of the number $I(t)$ of infective hosts over times $t\in[0,t']$ in terms of the number $\bar{I}_n=I(\tau_n+0)$ of infective hosts at equidistant times $\tau_n=n\tau$, for $n\in\{0,1,...,m\}$, with $\tau=m^{-1}t'$. Although it appears to be analytically intractable, the problem is closely related to the distribution of the total area between the sample paths of infectives in the continuous-time process ${\cal I}(t')=\{I(t): t\in [0,t']\}$ and its discrete-time counterpart $\bar{\cal I}^{(m)}=\{\bar{I}_n: n\in\{0,1,...,m\}\}$. Based on the effect of extreme values on the dynamics of ${\cal I}(t')$ and $\bar{\cal I}^{(m)}$, we conduct numerical results which show that, for any time interval of a predetermined length $t'>0$, it is generally possible to replace the SIS-model by a time-discretized version that is suitably selected by comparing the Hellinger distance between the corresponding extreme values distributions. We derive analytical expressions for key indexes --linked to the stationary numbers of infective hosts, the random length of an outbreak and the non-detection of an outbreak--, and we highlight the implications of their importance in the replacement of the original SIS-model by a certain time-discretized version. Our work complements to the work of Allen and Burgin ({\it Mathematical Biosciences}, {\bf 163,} (2000), 1--33), who define discrete-time SIS-models by assuming that at most one event (either an infection or a recovery) occurs in time steps of length $dt$, which is assumed to be sufficiently small.

17:15 to 17:40 Ravishankar. N Statistical modelling of Dengue incidences and climatic variables in India

This talk focuses on “Dengue”, which is considered as a one of the major public health threats in India. It aims to model the yearly state wise dengue incidences with respect to selected climatic variables namely annual rainfall and average temperature from the year 2008 to 2017 and identify hotspots of dengue incidence in states of India. Data required for the study were collected from www.indiastat.com. The expected dengue incidences were modelled using negative binomial regression with states as random effect and covariates as climatic variables and year. To identify the spatial clustering of dengue incidences, appropriate spatial clustering algorithm was used. To assess the spatial clustering of dengue incidences, Local Indicator of Spatial Association (LISA) was used. The hot spots for dengue incidence in states of India were determined using GeoDa/SatScan. From the study it was found that there is no association between the dengue incidences and climatic variables - average temperature and annual rainfall in India. Karnataka, Kerala and Tamilnadu were found to be the hotspots of dengue incidences

Tuesday, 02 July 2019
Time Speaker Title Resources
09:00 to 10:00 -- Discussions and Collaborations
10:00 to 10:50 Murad Banaji Some themes in modern chemical reaction network theory

Chemical reaction networks (CRNs) arise naturally as components of many physical and biological systems. Even when the elements of a system are not chemical in nature, the models which arise may be formally identical to CRN models. CRN theory focusses on making parameter-independent claims about CRNs. In other words, the goal is to describe behaviours of CRN models which arise as a consequence of the network structure and are largely independent of model details. The claims are generally about allowed or forbidden asymptotic behaviours. There are strands of work on conditions for multistationarity and oscillation; on the question of ”persistence”, namely whether some species can asymptotically ”run out” in a given CRN; on conditions for global stability of equilibria; and so forth. In many cases the results can naturally be implemented as algorithms, allowing for automated analysis of CRNs. The results from parameter-independent analysis form a natural precursor to more numerical and computational approaches needed for detailed analysis of CRN models.

Classical CRN theory focussed heavily on systems with mass action kinetics and gave rise to some powerful and useful theorems, particularly the deficiency zero and deficiency one theorems. In more recent times, the field has seen a resurgence of interest, with diverse mathematical tools being brought into play. This talk will outline some current themes in CRN theory, particularly focussed on how the presence or absence of particular subnetworks (”motifs”) influences allowed dynamical behaviours in ordinary differential equation models of CRNs. There are a number of results which take the form: ”a CRN containing no subnetworks of type X cannot display behaviour Y”. In the other direction we have results of the form: ”if a CRN contains a subnetwork of type X, then some model of this CRN admits behaviour Y”. The results are subtle and rely on appropriate notions of subnetwork, and assumptions about the class of kinetics. After a broad
introduction to the mathematical theory of CRNs, I’ll describe some of these results, and sketch some current challenges in this area.

10:50 to 11:20 -- Coffee break
11:20 to 12:10 Suman Chakrabarty Conformational free energy landscape of misfolding and aggregation in Prion proteins: A challenge for enhanced sampling techniques

There exists a toxic misfolded beta-sheet rich “scrapie” (PrPSc) isoform of Prion protein that induces misfolding in the healthy cellular form followed by fibril formation. Despite enormous efforts the structure of the scrapie remains elusive. Our goal is to use enhanced sampling techniques like Replica Exchange Molecular Dynamics (REMD) and metadynamics simulations to explore the complex conformational free energy landscape of Prion monomers and dimers to identify the structure of the scrapie form. Our study indicates that cellular form of Prion has a rather shallow free energy landscape with multiple low-lying metastable states that render the protein prone to misfolding. Also, we find that non-native hydrogen bonded interactions tend to stabilise these metastable states.

References:
• N. Chamachi and S. Chakrabarty, J. Phys. Chem. B 120, 7332 (2016)
• R. Singh, N. Chamachi, S. Chakrabarty and A. Mukherjee, J. Phys. Chem. B 121, 550 (2017)
• N. Chamachi and S. Chakrabarty, Biochemistry 56, 833 (2017)

12:10 to 13:00 Kavita Jain Adaptation slow down due to deleterious mutations

In a large population, the chance and the rate at which beneficial mutation spreads is influenced by the presence of superior beneficial mutations as well as deleterious mutations that, for example, can cause disease. I will introduce some basic evolutionary concepts pertinent to the discussion and present our recent theoretical results on how interference between deleterious and beneficial mutations decreases the rate of adaptation in asexual populations.

13:00 to 14:15 -- Lunch
14:15 to 15:05 Sutirth Dey Stabilizing Biological Populations: The Experimental Biologist’s Perspective

A large number of methods have been proposed to control the dynamics of an unstable system to a desired state. However, the predictions of these methods have rarely been tested on data from real, biological populations. Moreover, few studies have compared these methods against each other to understand their relative strengths and weaknesses. In this talk, I am going to present the contributions of our group in terms of addressing these lacunae using empirical time series from laboratory populations of Drosophila melanogaster. Specifically, I am going to show how different ecologically relevant stability properties can be altered through appropriate perturbations in the number of individuals. I am going to compare these empirical observations with simulation results from simple population dynamics models. I am also going to comment on how life-history parameters affect population dynamics, using perhaps the most comprehensive, biologically relevant model for laboratory populations of Drosophila melanogaster.

15:05 to 15:55 Nagasuma Chandra Identifying strategic target combinations in drug-resistant bacteria through systems modelling

The emergence of drug-resistant strains of M. tuberculosis poses a major threat to public health, warranting urgent attention to the problem of tackling antimicrobial resistance. A number of studies have sought to understand the causes of drug resistance, and have led to the identification of several mechanisms including mutations in key targets, upregulation of drug efflux pumps, etc. Yet, there is no understanding of which mechanisms are operative in a given condition, whether microbes explore multiple mechanisms simultaneously, or if such mechanisms are influenced by each other and lead to alterations in the cell in a synchronized manner. Towards this, an understanding of the global mechanisms leading to resistance becomes necessary. In this talk, I will describe different modelling approaches that we have adopted to address some of these questions and describe how we use them to identify strategic target combinations to tackle antimicrobial resistance.

15:15 to 16:25 -- Coffee break
16:25 to 17:15 Martín López-García A unified stochastic modelling framework for the spread of nosocomial infections

Over the last years, a number of stochastic models have been proposed for analysing the spread of nosocomial infections in hospital settings. These models often account for a number of factors governing the spread dynamics: spontaneous patient colonization, patient-staff contamination/colonization, environmental contamination, patient cohorting, or healthcare workers (HCWs) hand-washing compliance levels. For each model, tailor-designed methods are implemented in order to analyse the dynamics of the nosocomial outbreak, usually by means of studying quantities of interest such as the reproduction number of each agent in the hospital ward, which is usually computed by means of stochastic simulations or deterministic approximations. In this work, we propose a highly versatile unified stochastic modelling framework that can account for all these factors simultaneously, and analyse the reproduction number of each agent at the hospital ward during a nosocomial outbreak, in an exact and analytical way. By means of five representative case studies, we show how this unified modelling framework comprehends, as particular cases, many of the existing models in the literature. We implement various numerical studies via which we: i) highlight the importance of maintaining high hand-hygiene compliance levels by HCWs, ii) support infection control strategies including to improve environmental cleaning during an outbreak, and iii) show the potential of some HCWs to act as super-spreaders during nosocomial outbreaks.

This work is based on the publication

L´opez-Garc´ıa M, Kypraios T (2018) A unified stochastic modelling framework for the spread of nosocomial infections.
Journal of the Royal Society Interface, 15: 20180060.

17:15 to 17:40 Narmada Sambaturu Incorporating genetic heterogeneity into epidemic models for H1N1 influenza

Genetic differences contribute to variations in the immune response mounted by different individuals to a pathogen. Such differential response can influence the spread of infectious disease, indicating why such diseases impact some populations more than others. For a population with known HLA class-I allele frequency and for a given H1N1 viral strain, we classify individuals into sub-populations according to their level of susceptibility to infection. Our core hypothesis is that the susceptibility of a given individual to a disease such as H1N1 influenza is inversely proportional to the number of high affinity viral epitopes the individual can present. This number can be extracted from the HLA genetic profile of the individual. We use ethnicity-specific HLA class-I allele frequency data, together with genome sequences of various H1N1 viral strains, to obtain susceptibility sub-populations for 61 ethnicities and 81 viral strains isolated in 2009, as well as 85 strains isolated in other years. Our results show that HLA allele profiles which lead to a large spread in individual susceptibility values can act as a protective barrier against the spread of influenza. Several studies have reported that during the 2009 pandemic, indigenous ethnicities experienced more severe epidemics than their non-indigenous counterparts. Based on the observations during the 2009 pandemic, it has been suggested that aboriginal communities should be prioritised during vaccination. However, our results suggest that at least from the perspective of HLA alleles and downstream cytotoxic T lymphocyte response, this generalization does not hold true, and each influenza strain and each aboriginal community needs to be assessed independently. We also find that high risk alleles for one strain do not always correlate with severe epidemics in general. Using our model, it is possible to predict whether or not a new strain will cause a worse epidemic than a strain in the data set, within the constraints of the assumptions made. Predictions such as these could help optimise the deployment of resources when combating a new strain of influenza.

Wednesday, 03 July 2019
Time Speaker Title Resources
09:00 to 10:00 -- Discussions and Collaborations
10:00 to 10:50 Ganesh Bagler Computational Gastronomy: Leveraging food for better health through data-driven investigations

Cooking forms the core of our cultural identity other than being the basis of nutrition and health. Starting with a seemingly simple question, ‘Why do we eat what we eat?’, data-driven research conducted in our lab have led to interesting explorations of patterns in traditional recipes, their flavor composition, and health associations. Our investigations have revealed ‘culinary fingerprints’ of regional cuisines across the world, starting with the case study of Indian cuisine. Increasing availability of culinary data and the advent of computational methods for their scrutiny is dramatically changing the artistic outlook towards gastronomy. Application of data-driven strategies for investigating the gastronomic data (such as traditional recipes, molecular constituents of ingredients, percepts of flavor compounds, and health associations of food) has opened up exciting avenues giving rise to an all-new field of ‘Computational Gastronomy’. This emerging interdisciplinary science asks questions of culinary origin to seek their answers via compilation of culinary data and their analysis using methods of statistics, machine learning, natural language processing, pattern mining, and chemo-informatics. Along with complementary experimental studies, it has the potential to transform the food landscape by effectively leveraging data-driven food innovations for better health and nutrition. This talk will provide an overview of computational gastronomy research from our lab (Complex Systems Laboratory, IIIT-Delhi) with emphasis on data-driven investigations probing food and their health associations.

10:00 to 10:50 Supreet Saini Coordinated control of virulence factors in Salmonella enterica

Micro-organisms are constantly monitoring their surrounding environment and making important lifestyle decisions. This decision process is governed by large genetic networks that process the information leading to a phenotypic response from the cell. Using the food-borne pathogen, Salmonella, as a model organism, I will discuss how cells encode strategies in networks to optimally control cellular behavior.

Salmonella, on ingestion with contaminated food, swims in small intestine using propeller-like structures, flagella, on its surface. On reaching the site of infection, it assembles a hypodermic needle on its surface using which it injects proteins into host cells. These proteins cause a change in the host-cell shape leading to internalization of the bacterium. If the bacterium fails to get internalized, it assembles finger-like projections, fimbriae, on its surface to adhere to and persist in the intestine. How Salmonella dynamically regulates gene expression and assembly of these organelles is the focus of this study. I will show that the networks controlling genes necessary for flagella, needle, and fimbriae are designed so as to encode logic gates and limit expression to conditions optimum for infection. In addition, there is cross-talk between these three systems which serves to dynamically control the timing of activation and de-activation of these networks. Collectively, cells dynamically process information in genetic networks which ensures that the encoded products are produced at the correct locales, at the appropriate levels, and for the appropriate amount of time

10:50 to 12:10 -- Coffee Break + Poster session
12:10 to 13:00 Narendra Dixit Engineering the germinal center reaction for functional cure of HIV infection

Following an infection or vaccination, focused Darwinian selection in germinal centers preferentially selects B cells expressing surface receptors with increasing affinities for a target antigen. This process of affinity maturation is only rarely seen in HIV infected individuals to yield broadly neutralizing antibodies, which can control the infection and potentially prevent the progression to AIDS. In recent studies, passively administering such antibodies has been found, unexpectedly, to significantly improve this affinity maturation. In this talk, I will present a stochastic simulation model of the germinal center reaction in the presence of passively administered antibodies that explains this latter observation. We find that germinal centers are constrained by a quality-quantity trade-off. Increasing selection stringency improves the affinity of the selected B cells for antigen but compromises their numbers. Passively administered antibodies modulate this trade-off by altering the selection stringency in the germinal centers. Our simulations explain several independent experimental observations and present a framework to identify optimal passive immunization protocols. We predict that the optimal protocols would depend on the antigen level in the germinal centers, a correlate of pathogen load, using which the protocols could be personalized.

13:00 to 14:15 -- Lunch
14:15 to 15:05 Chaitanya A. Athale Effect of growth rate and population density on bacterial cell size

Cell size variation in a population of genetically identical Escherichia coli has been observed in the past. While the regulation of cell size involves the convergence of multiple genes and proteins, they typically converge on cell division and DNA replication and segregation. We sought to answer how the readout of such factors in terms of growth rate affect cell size and whether cell density has any role to play. Here, we have developed a model of bacterial DNA replication coupled to cell division in the context of a logistically growing population. We have chosen to model the growth of E. coli based on the BCD-Model where DNA replication is modeled as the stochastic dynamics of replication fork (RF) progression. RFs are modeled to exist in two states- stalled and recovered and transition stochastically between the states. Stalled RFs or incomplete DNA replication results in aberrant cell division. With these simple assumptions we find cell-size variability of cultures depends strongly on growth rate and not population size or growth phase in the context of a logistic model. This is consistent with experimental data [1]. To test the model, we compare model perturbations to experimental cell length measurements of E. coli recA, sulA and slmA, since these genes act as sensors and actuators of RecA and nucleoid occlusion coupling to cell division [2]. Our model explores the role of replication stochasticity and in future work we are attempting to understand potential functional roles for cell size diversity in spatial spread.

References:
[1] Gangan M.S. and Athale C.A. (2017) Threshold Effect of Growth Rate on Population Variability of Escherichia coli Cell Lengths Roy. Soc. open science DOI: 10.1098/rsos.160417

[2] Athale C.A. Effect of Replication Fork Dynamics on Escherichia coli Cell Size [arXiv: 1601.02240]

15:05 to 15:55 Ruy Ribeiro HIV-1 viral kinetics during treatment with multiple drug classes

Modeling of viral dynamics has contributed to our understanding of the biology of multiple viruses. In particular, analyzing the kinetics of HIV-1 decay during antiretroviral therapy has elucidated details of viral infection, and parameters of virus and infected cell turnover, as well as the cell populations contributing to viral production. Antivirals from the integrase inhibitor class give rise to a pattern of HIV decay that is different from what had been observed before with other antiretrovirals. We modeled the effect of integrase inhibitors in short- and long-lived infected cells and fitted the model to data obtained from participants treated with different drug classes, including or not integrase inhibitors. This new model explains and quantifies the three phases of HIV-1 RNA decay with an integrase inhibitor vs. the two phases observed in therapies without it. In this way, the model provides a mechanistic description of viral infection that parsimoniously explains the kinetics of viral load decline under multiple classes of antiretrovirals.

15:55 to 16:25 -- Coffee break
16:25 to 17:15 Prem Jagadeesh Challenges in Identification of Quantitative System Pharmacology (QSP) models

Quantitative systems pharmacology (QSP) is a subset of pharmacometrics, which concerns the quantitative analysis of dynamic interactions between drugs and a biological system, with the aim to understanding the system as a whole. QSP models are often detailed mechanistic models, which have a large number of unknown parameters that have to be estimated from observed data. Since most biological processes are non-linear in nature, the identification of such highly parametrised non-linear models poses multiple challenges in the identification exercise. Sloppiness is one of the long discussed topics in the identification of biological processes. Sloppiness essentially quantifies anisotropy of model output to infinitesimal perturbations in the parameter space. Previous studies have shown that, for sloppy models, identification from data will result in extremely poor parameter estimates but reasonably good predictions. Sloppiness is placed somewhere in-between strict identifiability and loss of identifiability. Sloppiness seems to be closely connected to the concept of identifiability, but the exact relationship has hitherto not been elucidated. In this work, we have revisited the definition for sloppiness in the line of identifiability. We elucidate with simple toy examples, the exact roles of model structure, data and cost function in sloppiness and finally arrive at the necessary condition for sloppiness as the model structure. We have extended the concept of equality of model structures in discrete time systems to continuous time systems and delineated the difference between sloppiness and loss of identifiability due to data in the lens of predictions.

Thursday, 04 July 2019
Time Speaker Title Resources
09:10 to 10:00 Saumyadipta Pyne Hierarchical Modeling of High-dimensional Human Immuno-phenotypic Diversity

Human immuno-phenome can provide critical insights into disease dynamics and clinical classification. A major obstacle in the efforts of precise characterization of the human immuno-phenome is the immense diversity therein. Immuno-phenotype data may be generated in high-resolution at the level of single cells, and often they present challenges for current analytical methods due to their high dimensionality, large number of observations, as well as complex distributional features such as multimodality, asymmetry, and other non-normal characteristics. We developed computational frameworks specifically for fast and automatic modeling and identification of different cell populations, their hierarchical structures and inter-relationships under different biological conditions.

10:00 to 10:50 Martin CJ Bootsma Efficacy of hand hygiene, cohorting of health care workers and patients and isolation

Prevention of pathogen transmission in hospital settings can be pursued by measures targeting a single pathogen (vertical intervention) or targeting more pathogens simultaneously (horizontal intervention).  Hand hygiene and universal cohorting of patients and health care workers (HCWs) are horizontal interventions, whereas cohorting of patients with known colonization, or improved hand hygiene by HCWs after contact with known colonized patients (isolation) are vertical interventions. We develop a stochastic modeling framework, based on the Master equation, to quantify the effects of different cohorting schemes, isolation and hand hygiene improvement on the prevalence of colonization with microorganisms in a small hospital unit with two types of HCWs, nurses (that can be cohorted) and physicians (that cannot be cohorted). Hand hygiene by HCWs is modeled as a chance of hand decontamination between two patient contacts. Hand hygiene adherence is assumed to be 43% for physicians and 59% for nurses and 31% of the patient contacts are due to physicians and 69% due to nurses.
We consider three cohorting schemes; (1) no cohorting (mass-action assumption); (2) 1st-order cohorting (each patient has a preferred nurse and, if unavailable, other nurses are equally likely to provide care); and (3) higher-order cohorting 1st-order cohorting incorporating also cohorting when first assigned nurses are not available) and we consider both horizontal and vertical cohorting and we vary the per admission reproduction number from 0.5 to 2 and the admission prevalence from 1% to 10%.
Hand hygiene improvement is more efficient than previous modeling studies suggest, as we take multiple failures in hand decontamination by HCWs in a row into account. When 80% of the contact of nurses are with patients in their own cohort, 1st-order horizontal cohorting can reduce the number of acquisitions with 17% (R0=0.5, admission prevalence f=0.1 to 34% for R0=2 and f =0.01. In the absence of cohorting hand hygiene adherence of both nurses and physicians should improve by 16%, increasing the hand hygiene adherence from 0.59 to 0.66 for nurses and from 0.43 to 0.52 for physicians) and 18%, respectively, to realize the same prevalence reductions.  Higher-order cohorting can further augment effectiveness. The ultimate limiting factor are uncohorted physicians. Vertical cohorting is even more effective, and negative effects on the prevalence of pathogens for which vertical cohorting is not applied are very limited. A targeted hand hygiene improvement of 20%, either between patient contacts in different cohorts or after contact with isolated patients reduces the number of acquisitions with 14%-63% and 8%-37%.
We provide a general and flexible methodology to investigate the impact of transmission prevention measures. Our findings demonstrate that hand hygiene improvement, cohorting and isolation are powerful tools to prevent pathogen transmission in healthcare settings.
Given the well-known inverse relationship between workload of HCWs and hand hygiene compliance, our findings support more attention for staffing levels and cohorting structures as a measure to interrupt nosocomial transmission of pathogens.

10:50 to 11:20 -- Coffee break
11:20 to 12:10 Emma McBryde Mathematical and Statistical modelling in the fight to control Tuberculosis

Background: Tuberculosis (TB) continues to be the worlds greatest infectious diseases killer, with nearly 2million people per year dying from the disease. Many features of tuberculosis make it challenging to detect, understand and control including its long latent period, masked presentation particularly in children, and the heterogeneity of all aspects of the disease. Missing cases are a major reason TB is not well controlled today, as it is estimated that one in three cases of TB go undiagnosed.
Unlike many other diseases of major public health threat, HIV and malaria for example, TB is not well characterised in terms of infectiousness over the natural history of infection, and even in its latency period. Although different structures are used in modern TB models to simulate TB latency, it remains unclear whether they are all capable of reproducing the particular activation dynamics empirically observed. Statistical and mathematical models combined can be used to challenge the models with data and refine parameters to provide insights into disease dynamics and also potentially into the disease itself.
Aims: I will present work from my group on challenging model parameters with data specifically re-estimating the latent period of tuberculosis for different age groups and estimating death from tuberculosis by reanalysing the natural history data from the pre-chemotherapy era. I will then go on to show why getting these parameters matters from a disease control perspective.
Latency: We then fitted six different models (from the literature) to the activation dynamics observed from 1352 infected contacts diagnosed in low incidence countries (Australia and Netherlands) to obtain parameter estimates. Only models incorporating two latency compartments were capable of reproducing the activation dynamics empirically observed. Additionally, when two successive latency compartments are used, the first period should have a duration that is much shorter than that used in previous studies. Younger age categories showed a markedly higher early activation parameter. Models require at least two latency compartments and age-stratification in order to accurately replicate the dynamics of TB reactivation from latency. I will show why it is important to consider this in the fight against TB.
Mortality: Models of TB that use natural history parameters for mortality usually estimate death rates which are far in excess of those estimated by WHO and National TB control programs. The discrepancy is due to the death from cases missing from the national TB control program. It is difficult to reconcile data on mortality, missing cases and death rates from TB. I will present model results which suggest a large number of deaths from TB go unreported.

12:10 to 13:00 Chetan Gadgil Use of mathematical models for simulating drug efficacy and explaining unintuitive experimental observations

Through two examples, I will show the utility of mathematical models in suggesting explanations for unintuitive observations. I will also show how existing models can be used to suggest potential treatment strategies.
MicroRNAs bind to their target mRNAs and decrease the level of the translated product. However, several experimental studies conclude that for some miRNA-target pairs the target protein levels are positively correlated with miRNA levels. Through a mathematical model of the reactions governing miRNA and mRNA formation, interaction, and catalytic protein production, we show that the apparent positive regulation of targets by miRNA could be an artefact resulting from competition for mRNA (Nyayanit and Gadgil, RNA, 2015). The cKit/SCF pathway is known to regulate melanogenesis. It is also affected by several cancer drugs, including Imatinib, resulting in hypo-pigmentation as a side-effect in 40% of CML patients treated with Imatinib. There is also a subset of patients for whom the side effect is qualitatively opposite, i.e. they get hyper-pigmented on treatment with Imatinib. Using a simple mathematical model for the cKit pathway, we show that even for the same pathway topology and the same treatment effect, the net effect of inhibiting cKit on pigmentation may be qualitatively different. We will present unpublished results showing that a dimensionless number can predict whether pathway inhibition will lead to hyperpigmentation or hypopigmentation.
Systemically treating a disease with a drug results in its distribution in various organs. For diseases with multi-organ prevalence, successful treatment in one organ may not necessarily imply efficacy in other organs. Using a physiologically based pharmacokinetic model for anti-tuberculosis medication, I show that the standard treatment regime is unlikely to be efficacious in treating extra-pulmonary TB of organs such as the bone. I will describe planned work on combining a PBPK model of drug distribution with a model of MTb growth and evolution of resistant strains, and describe the use of such models to simulate treatment strategies.

13:00 to 14:15 -- Lunch
Friday, 05 July 2019
Time Speaker Title Resources
09:00 to 10:00 -- Discussions and Collaborations
10:00 to 10:50 Ganesh Viswanathan Phenotype switching during Tumor Necrosis Factor alpha signaling

Tumor Necrosis Factor alpha (TNFα) is a pleiotropic cytokine involved in phenotypic decisions such as apoptotic/necrotic death, proliferation. Aberrant TNFα signaling is implicated in numerous pathological conditions such as neurological disorders, autoimmune diseases, cancer. Designing therapeutic strategies to modulate these conditions require insights into the mechanisms governing context-specific phenotypic response to TNFα. Signal transduction culminating in such responses is orchestrated by underlying molecular network of nodes (such as proteins, genes) interconnected by edges (such as protein-protein, protein-gene interactions). Deciphering mechanism(s) and identifying nodes/interactions controlling specific phenotypic response requires a holistic approach involving systematic analyses of the TNFα signaling network. In this talk, a combination of functionally consistent topological analysis and boolean dynamics simulations implemented on a well-annotated comprehensive TNFα signaling network to find targets for phenotype switching (from proliferation to apoptosis or vice-versa) will be presented. In particular, a graph-theory based network dimensionality reduction via modularization and the robustness properties will be used for finding potential target candidates. Systematic attractor analysis of the state transition graph of the network will be used for assessing the extent of phenotype switching that can be achieved when tese candidates are knocked-out.

10:50 to 11:20 -- Coffee break
11:20 to 12:10 Vinod PK Network and trajectory inference approaches to understand the behavior of pancreatic cells in healthy and Type 2 diabetes

The advent of single-cell RNA sequencing (scRNA-seq) technique has generated valuable resource on islet biology and type 2 diabetes (T2D). This provides an opportunity to understand the different cell types/states at both the network and individual gene expression levels. In this study, we inferred the gene regulatory networks (GRNs) of pancreatic cells from available scRNA-seq data in healthy and T2D using single-cell regulatory network inference and clustering workflow. Clustering of cells based on GRNs identifies endocrine and exocrine cells and multiple stable cell states in each alpha, beta and ductal cells. The phenotypic variations in cell states due to obesity and T2D are indistinguishable. Therefore, the trajectory of cells in pseudotime was constructed based on the cell-type-specific gene expression. The analysis shows that continuous spectrum of cell states exists with phenotypic-dependent branching and donor cell-cell variability in endocrine and exocrine cell types. We identified the genes that give rise to bifurcation in the trajectory. Our analysis demonstrates that the heterogeneity in pancreatic cells can be characterized at the network- and gene-level with the help of scRNA-Seq data.

12:10 to 13:00 Sarika Jalan PEV localization in complex Networks: Relation with the disease spreading

“The term localization is used in physics to describe situations where the strength of interactions between different parts of a system decay rapidly with the distance: in other words, correlations are short-ranged. The opposite of localization is delocalization: a function is delocalized if its values are nonnegligible on an extended region. In other words, if long-range interactions are important, a system is delocalized. Localization (or lack of it) is of special importance in quantum chemistry and solid state physics since the properties of molecules, and the behavior of materials are strongly dependent on the presence (or absence) of localization” [1].

In complex networks, localization behavior of PEV and its corresponding eigenvalue of the networks’ adjacency matrices is known to provide an understanding of local and global structural as well as the dynamical behavior of the underlying systems [2, 3, 4, 5]. Importantly, the PEV of an adjacency matrix approximate the steady state vector of many linear dynamical systems which includes epidemic spreading model, RNA neutral networks model, rumor spreading models, brain network dynamical model [2]. The PEV entries interpret the contribution of nodes in such linear dynamical processes in the steady state. For example, PEV entries will predict which nodes are more likely to be affected during a disease spread. A network structure having localized PEV indicates that a few nodes contribute more in the dynamical process, and the rest of them have less contribution irrespective of the system size. Similarly, for a delocalized PEV, all the nodes have the same amount of contribution to a similar dynamical system. In the context of disease spread, if the PEV of the network’s adjacency matrix is localized, in the vicinity of the epidemic threshold for susceptible-infected-susceptible (SIS) model, the disease infects a small number of vertices and spreading process becomes slow. As a result, it requires a larger epidemic threshold to spread the disease over the network. Further, the localization of PEV is related to detect criticality in the brain network dynamics [6]. Moreover, localized eigenvectors are successful in the identification of microscopic functional units in the neural networks [5, 7]. Furthermore, bistable activities of signaling in the biological networks have also been examined through the localization of PEV of the corresponding adjacency matrices [8].

One key factor of our interest is to understand the properties of networks which may help in spreading or restricting perturbation in networks. For instance, during a disease outbreak or cell signaling propagation, one will be interested in knowing if the disease or cell signal will be pandemic or will be localized to a smaller section of the network. Similarly, one may be interested in spreading a piece of particular information, for instance, awareness of vaccination at the time of disease outbreak, or may wish to restrict or localize a perturbation like signal propagation. Since the perturbations in a complex system represented by networks can be well explained through PEV of the underlying adjacency matrices, we can mathematically frame the following question which we are addressing in the current work. Given a network’s parameter, the number of nodes (N) and connections (M), what network structure will correspond to the most localized PEV.  

To get a connected network which has the most localized PEV, we formulate an optimization problem through network evolution with edge rewiring scheme. Given an input graph G  with N  vertices, M  edges and a function ζ, we want to compute the maximum possible value of ζ(G) over all the simple, connected, undirected, and unweighted graph G. Thus, we are maximizing the objective function ζ(G) = IPR(x1) = (x1)1+ (x1)24  +...+ (x1)Nsubject to the constraints that (x1)12 + (x1)22 +...+ (x1)N2  = 1, and 0 < (x1)i < 1. Starting from a random network as an initial network, we achieve an optimized network structure having highly localized PEV through edge rewiring based method. The optimized structure consists of two subgraphs connected via a node. In addition, we provide a complete theoretical understanding of the emergence of such structures as a consequence of PEV localization without performing the network evolution process. To demonstrate the efficiency of these artificially constructed network structures having localized PEV for a dynamic process, we use the standard susceptible-infected-susceptible (SIS) disease spreading model [9]. We reveal that for the initial network, within short time disease infects a large number of vertices, whereas, for the optimized network, there exist very few vertices which get infected.

The results and framework provide a promising avenue to address concerns of various practical applications. For example, in many cases, it is not possible to perform an extensive rewiring the entire network, and instead, a small subset is available to perform any changes. Second, instead of rewiring, sometimes it is more realistic to add a few edges to construct a weighted network. Again the same question one can raise that how a small sub-structure is required, or few edges are added such that the final structure is highly localized.

References

[1] M. Benzi, Localization in Matrix Computations: Theory and Applications, Springer International Publishing AG, (2016).
[2] J. Aguirre, D. Papo, J. M. Buld ́u, Successful strategies for competing networks, Nat. Phys. 2013, 9:230.
[3] C. Castellano, R. Pastor-Satorras, Topological determinants of complex networks spectral properties: structural and dynamical effects, Phys. Rev. X 2017; 7:041024.
[4] P. Pradhan, A. Yadav, S. K. Dwivedi, and S. Jalan, Optimized evolution of networks for principal eigenvector localization, Phys. Rev. E 96 , 022312 (2017).
[5] R. Chaudhuri, A. Bernacchia, and X. Wang, A diversity of localized timescales in network activity, eLIFE 3, e01239 (2014).
[6] P. Moretti and M. A. Mu ̃noz, Griffiths phases and the stretching of criticality in brain networks, Nat. Commun. 4, 2521 (2013).
[7] M. T. Schaub, Y. N. Billeh, C. A. Anastassiou, C. Koch, and M. Barahona, Emergence of slow-switching assemblies in structured neuronal networks, PLoS Comput. Biol. 11, e1004196 (2015).
[8] G. Hernandez-Hernandez, J. Myers, E. Alvarez-Lacalle, and Y. Shiferaw, Nonlinear signaling on biological networks: The role of stochasticity and spectral clustering, Phys. Rev. E 95, 032313 (2017).
[9] A. V. Goltsev, S. N. Dorogovtsev, J. G. Oliveira, J. F. F. Mendes, Localization and Spreading of Diseases in Complex Networks, Phys. Rev. Lett. 2012; 109:128702.

13:00 to 14:15 -- Lunch
14:15 to 15:05 Karthik Raman Unraveling microbial interactions in the gut microbiome associated with antibiotic recovery

The human gut microbiome is a complex community of trillions of organisms, which suffers collateral damage post antibiotic treatment, but recovers subsequently. A recent study has elucidated the pivotal role played by twenty key species of bacteria in the gut, that enable microbiome recovery post antibiotic treatment. In this study, using computational tools to interrogate community metabolic networks, we unravel the dependencies that exist between these twenty microbes in the gut.

Using our previously developed graph-based algorithm, MetQuest, we enumerate a number of biosynthetic pathways, particularly involved in amino acid biosynthesis, and spanning across a pair of organisms. We compute a Metabolic Support Index, which captures the extent of support/interaction between a pair of microbes, based on the incremental change in metabolic capabilities achieved in a community vis-`a-vis the individual organisms. We show that organisms belonging to Firmicutes phylum benefit via metabolic interactions with those of Bacteroidetes phylum in minimal glucose as well as high fibre diet conditions. Our results also suggest that many of the metabolic interactions between the microbes are environment-dependent. We also construct a microbial association network, based on the improvement in the spectrum of an organism’s amino acid synthesis capabilities in the presence of another organism in the community. Overall, we present a repertoire of techniques to analyse interactions in generic microbial communities. Using these techniques, we predict several putative interactions in the gut microbiome, and compare against experimental observations reported in literature where available. Our predictions also provide testable hypotheses to guide further experimental validations.

15:05 to 15:55 Mohit Kumar Jolly Dynamical systems biology of cancer metastasis

Metastasis – the spread of cancer cells from one organ to another – causes above 90% of all cancer-related deaths. Despite extensive ongoing efforts in cancer genomics, no unique genetic or mutational signature has emerged for metastasis. However, a hallmark that has been observed in metastasis is adaptability or phenotypic plasticity – the ability of a cell to reversibly switch among different phenotypes in response to various internal or external stimuli. This talk will describe how mechanism-based mathematical models can help (a) identify how cancer cells can leverage this plasticity to drive cancer metastasis, (b) interpret confounding experimental and clinical data, and (c) guide the next set of crucial in vitro and in vivo experiments. Collectively, this work highlights how an iterative crosstalk between mathematical modeling and experiments can both generate novel insights into the emergent dynamics of cellular plasticity and uncover previously unknown accelerators of metastasis.

15:55 to 16:25 -- Coffee break
16:25 to 17:15 Sarah Russell Divergence in the T cell response is manifest from the fifth generation, but imprinted in the nave founder cell

Surviving an infection or cancer often depends upon the extent and speed with which specific CD8+ T cells are amplified. It is not yet clear what drives the variation in the T cell response. We image the entire T cell response at single cell resolution to explore the contribution of each cell to the clonal response, with three major findings. First, the proliferation rate up to the fourth generation is tightly constrained, and does not contribute to clonal dominance. In contrast, divergence in cell cycle times and cell death from the fifth generation determines which clones will dominate. Second, characteristics of cell cycle and death are predominantly determined by the identity of the founder cell. Third, the size of the founder cell, prior to antigen presentation, is a strong predictor of the clonal response.

17:15 to 17:40 Sruthi C.K. Deep2Full: Computational strategies to make complementary predictions of the effects of the massively parallel disease or antibiotic resistance causing mutations

Mutations in protein can affect the structure and function of the protein leading to changes in the cellular fitness and consequently to several diseases such as cancers. Mutagenesis experiments have been used to understand this genotype to phenotype mapping. Owing to the developments in the sequencing technologies large scale mutagenesis experiments such as deep mutational scan has become possible probing phenotypic consequences of thousands variants of a protein. But often these maps are incomplete, missing the data of a few hundreds of mutants. It has been shown that assay sensitive computational models can be developed using machine learning techniques to impute the scores of missing mutants by training on partial data available from deep mutational scan. This possibility of reliable prediction gives the option of performing experiments for a smaller set of variants using the data of which computational models could be built. Exploring this direction, we define and compare the prediction qualities of models generated by training on mutations chosen with different strategies. As expected, using 85% for the training data gave good results, and the strategy of decreasing the training set was satisfactory up to about 15% data. We also evaluated the strategies of choosing substitutions by alanine, asparagine and histidine at all positions and of randomly selecting mutants of the same number and found them comparable.

Saturday, 06 July 2019
Time Speaker Title Resources
09:00 to 10:00 -- Discussions and Collaborations
10:00 to 10:50 Katharine Best Modelling dose dependent immune responses in acute Zika infection

The key mechanisms of immune control of acute Zika virus infection are not fully understood, and furthering our knowledge of the within host dynamics of Zika virus will be important to the development of effective antiviral strategies. The majority of within host dynamics research has been performed in non-human primate (NHP) models of Zika infection, and understanding the role of inoculum dose is an important component in being able to translate results from a controlled experimental infection to a natural infection. Here we use mathematical modelling to analyze the within host dynamics of Zika virus in NHPs after infection at different inoculum doses. We find strong evidence for dose dependent innate immune control of plasma viral load.

10:50 to 11:20 -- Coffee break
11:20 to 12:10 Laura Liao Mathematical modelling of ebola virus infection in vitro

Mathematical modellers have been able to fully characterize the replication kinetics of influenza A virus infections in vitro. These results have provided insight on the impact of a particular antiviral resistance mutation, and explained strain differences between avian and human influenza A virus. In their approach, the combination of a simple mathematical model and a key set of experiments was essential to the analysis. We discuss their approach, and apply a similar analysis to ebola virus infection in vitro for which preliminary results will be presented.

12:10 to 13:00 Carmen Molina-Paris Stochastic descriptors to study the fate of naive T cell clonotypes in the periphery

The population of naive T cells in the periphery is best described by determining both its T cell receptor diversity, or number of clonotypes, and the sizes of its clonal subsets. In this talk, we make use of a previously introduced mathematical model of naive T cell homeostasis, to study the fate and potential of naive T cell clonotypes in the periphery. This is achieved by the introduction of several new stochastic descriptors for a given naive T cell clonotype, such as its maximum clonal size, the time to reach this maximum, the number of proliferation events required to reach this maximum, the rate of contraction of the clonotype during its way to extinction, as well as the time to a given number of proliferation events. I will show that two fates can be identified for the dynamics of the clonotype: extinction in the short-term if the clonotype experiences too hostile a peripheral environment, or establishment in the periphery in the long-term. In this second case the probability mass function for the maximum clonal size is bimodal, with one mode near one and the other mode far away from it. Our model also indicates that the fate of a recent thymic emigrant (RTE) during its journey in the periphery has a clear stochastic component, where the probability of extinction cannot be neglected, even in a friendly but competitive environment. On the other hand, a greater deterministic behaviour can be expected in the potential size of the clonotype seeded by the RTE in the long-term, once it escapes extinction.

13:00 to 14:15 -- Lunch
Monday, 08 July 2019
Time Speaker Title Resources
09:00 to 10:00 -- Discussions and Collaborations
10:00 to 11:00 Carmen Molina-París Adventures in mathematical immunology
11:00 to 11:30 -- Coffee break
11:30 to 12:30 Ruy Ribeiro Introduction to data analyses with R (Tutorial Session 1)

This hands-on section will provide an introduction to R (a language and environment for statistical computing), with specific applications to biological data analyses. We will start with an overview of this system, and its basic operations and data structures. We will build on these to showcase important features such as scripting of functions, statistical analyses and graphics. I will also present several interfaces and the power of packages.

It is expected that the students will have installed R and Rstudio on their computers, to work on practical examples.

Download R at https://cran.r-project.org/
Download RStudio at https://www.rstudio.com/

12:30 to 13:30 -- Lunch
13:30 to 14:30 Ruy Ribeiro Introduction to data analyses with R (Tutorial Session 1)

This hands-on section will provide an introduction to R (a language and environment for statistical computing), with specific applications to biological data analyses. We will start with an overview of this system, and its basic operations and data structures. We will build on these to showcase important features such as scripting of functions, statistical analyses and graphics. I will also present several interfaces and the power of packages.

It is expected that the students will have installed R and Rstudio on their computers, to work on practical examples.

Download R at https://cran.r-project.org/
Download RStudio at https://www.rstudio.com/

14:30 to 15:30 Ruy Ribeiro Introduction to data analyses with R (Tutorial Session 1)

This hands-on section will provide an introduction to R (a language and environment for statistical computing), with specific applications to biological data analyses. We will start with an overview of this system, and its basic operations and data structures. We will build on these to showcase important features such as scripting of functions, statistical analyses and graphics. I will also present several interfaces and the power of packages.

It is expected that the students will have installed R and Rstudio on their computers, to work on practical examples.

Download R at https://cran.r-project.org/
Download RStudio at https://www.rstudio.com/

15:30 to 16:00 -- Coffee break
16:00 to 17:00 Ruy Ribeiro Introduction to data analyses with R (Tutorial Session 1)

This hands-on section will provide an introduction to R (a language and environment for statistical computing), with specific applications to biological data analyses. We will start with an overview of this system, and its basic operations and data structures. We will build on these to showcase important features such as scripting of functions, statistical analyses and graphics. I will also present several interfaces and the power of packages.

It is expected that the students will have installed R and Rstudio on their computers, to work on practical examples.

Download R at https://cran.r-project.org/
Download RStudio at https://www.rstudio.com/

Tuesday, 09 July 2019
Time Speaker Title Resources
09:00 to 10:00 -- Discussions and Collaborations
10:00 to 11:00 David Sinclair Agent-based modeling with FRED (Tutorial Session 2)

FRED (a Framework for Reconstructing Epidemiological Dynamics) is an agent-based model platform that uses synthetic populations that capture the demographic and geographically heterogeneous characteristics of actual populations, including neighborhood, household, school and workplace social networks. FRED populations are currently available for the United States and Telangana, India. FRED is a flexible system that can be used to model diverse situations, including infectious diseases, chronic conditions, demographic changes at the population level, and the effects of human behavior. FRED can be used to predict outcomes of an epidemic, such as epidemic curves and disease incidence, and is a tool for evaluating interventions, such as school closure, social distancing and changes in vaccination behavior.
Workshop participants will become familiar with the FRED platform, learning the basics of creating, running and analyzing agent-based simulations. The workshop will use FRED Web, an online interface to FRED. FRED Web is designed to facilitate model development and simulation, with a graphic user interface to guide model development.

11:00 to 11:30 -- Coffee break
11:30 to 12:30 David Sinclair Agent-based modeling with FRED (Tutorial Session 2)

FRED (a Framework for Reconstructing Epidemiological Dynamics) is an agent-based model platform that uses synthetic populations that capture the demographic and geographically heterogeneous characteristics of actual populations, including neighborhood, household, school and workplace social networks. FRED populations are currently available for the United States and Telangana, India. FRED is a flexible system that can be used to model diverse situations, including infectious diseases, chronic conditions, demographic changes at the population level, and the effects of human behavior. FRED can be used to predict outcomes of an epidemic, such as epidemic curves and disease incidence, and is a tool for evaluating interventions, such as school closure, social distancing and changes in vaccination behavior.
Workshop participants will become familiar with the FRED platform, learning the basics of creating, running and analyzing agent-based simulations. The workshop will use FRED Web, an online interface to FRED. FRED Web is designed to facilitate model development and simulation, with a graphic user interface to guide model development.

12:30 to 13:30 -- Lunch
13:30 to 14:30 David Sinclair Agent-based modeling with FRED (Tutorial Session 2)

FRED (a Framework for Reconstructing Epidemiological Dynamics) is an agent-based model platform that uses synthetic populations that capture the demographic and geographically heterogeneous characteristics of actual populations, including neighborhood, household, school and workplace social networks. FRED populations are currently available for the United States and Telangana, India. FRED is a flexible system that can be used to model diverse situations, including infectious diseases, chronic conditions, demographic changes at the population level, and the effects of human behavior. FRED can be used to predict outcomes of an epidemic, such as epidemic curves and disease incidence, and is a tool for evaluating interventions, such as school closure, social distancing and changes in vaccination behavior.
Workshop participants will become familiar with the FRED platform, learning the basics of creating, running and analyzing agent-based simulations. The workshop will use FRED Web, an online interface to FRED. FRED Web is designed to facilitate model development and simulation, with a graphic user interface to guide model development.

14:30 to 15:30 David Sinclair Agent-based modeling with FRED (Tutorial Session 2)

FRED (a Framework for Reconstructing Epidemiological Dynamics) is an agent-based model platform that uses synthetic populations that capture the demographic and geographically heterogeneous characteristics of actual populations, including neighborhood, household, school and workplace social networks. FRED populations are currently available for the United States and Telangana, India. FRED is a flexible system that can be used to model diverse situations, including infectious diseases, chronic conditions, demographic changes at the population level, and the effects of human behavior. FRED can be used to predict outcomes of an epidemic, such as epidemic curves and disease incidence, and is a tool for evaluating interventions, such as school closure, social distancing and changes in vaccination behavior.
Workshop participants will become familiar with the FRED platform, learning the basics of creating, running and analyzing agent-based simulations. The workshop will use FRED Web, an online interface to FRED. FRED Web is designed to facilitate model development and simulation, with a graphic user interface to guide model development.

15:30 to 16:00 -- Coffee break
Wednesday, 10 July 2019
Time Speaker Title Resources
09:00 to 10:00 Mohit Kumar Jolly Communicating science to non-experts (Tutorial Session 3)

The need to acquire the skills to communicate one’s research to non-experts is being increasingly recognized by all stakeholders involved in research and development - funding agencies, journals, academic universities, as well as industries. Learning these skills is absolutely crucial today not only because interdisciplinary research where scientists trained in multiple disciplines collaborate based on tools and/or concepts is becoming the norm of the day, but also because many journals/granting agencies require authors to submit a ‘Significance Statement’ that should be understood by an ‘educated layman’.
By the end of this tutorial session that involves multiple interactive exercises such as improvisation and roleplaying, students will be able to:
1. Realize the jargon in the description of their research
2. Learn how to pitch their research in an engaging and exciting manner to a diverse audience
3. Consider the various need of a large audience

10:00 to 11:00 Mohit Kumar Jolly Communicating science to non-experts (Tutorial Session 3)

The need to acquire the skills to communicate one’s research to non-experts is being increasingly recognized by all stakeholders involved in research and development - funding agencies, journals, academic universities, as well as industries. Learning these skills is absolutely crucial today not only because interdisciplinary research where scientists trained in multiple disciplines collaborate based on tools and/or concepts is becoming the norm of the day, but also because many journals/granting agencies require authors to submit a ‘Significance Statement’ that should be understood by an ‘educated layman’.
By the end of this tutorial session that involves multiple interactive exercises such as improvisation and roleplaying, students will be able to:
1. Realize the jargon in the description of their research
2. Learn how to pitch their research in an engaging and exciting manner to a diverse audience
3. Consider the various need of a large audience

11:00 to 11:30 -- Coffee break
11:30 to 12:30 Mohit Kumar Jolly Communicating science to non-experts (Tutorial Session 3)

The need to acquire the skills to communicate one’s research to non-experts is being increasingly recognized by all stakeholders involved in research and development - funding agencies, journals, academic universities, as well as industries. Learning these skills is absolutely crucial today not only because interdisciplinary research where scientists trained in multiple disciplines collaborate based on tools and/or concepts is becoming the norm of the day, but also because many journals/granting agencies require authors to submit a ‘Significance Statement’ that should be understood by an ‘educated layman’.
By the end of this tutorial session that involves multiple interactive exercises such as improvisation and roleplaying, students will be able to:
1. Realize the jargon in the description of their research
2. Learn how to pitch their research in an engaging and exciting manner to a diverse audience
3. Consider the various need of a large audience

12:30 to 13:30 -- Lunch
13:30 to 14:30 Mohit Kumar Jolly Communicating science to non-experts (Tutorial Session 3)

The need to acquire the skills to communicate one’s research to non-experts is being increasingly recognized by all stakeholders involved in research and development - funding agencies, journals, academic universities, as well as industries. Learning these skills is absolutely crucial today not only because interdisciplinary research where scientists trained in multiple disciplines collaborate based on tools and/or concepts is becoming the norm of the day, but also because many journals/granting agencies require authors to submit a ‘Significance Statement’ that should be understood by an ‘educated layman’.
By the end of this tutorial session that involves multiple interactive exercises such as improvisation and roleplaying, students will be able to:
1. Realize the jargon in the description of their research
2. Learn how to pitch their research in an engaging and exciting manner to a diverse audience
3. Consider the various need of a large audience

14:30 to 15:30 Jonathan Carruthers Stochastic modelling of processes in Biology I (Tutorial Session 4)

In this workshop, we introduce approaches to study biological populations using stochastic processes. Through deriving the Kolmogorov equations, we can form systems of ODEs that describe the time evolution of the transition probabilities, the probability of the stochastic process being in a particular state at a particular time. For some processes, analytic solutions to these equations can be obtained, however, when this is not possible the Kolmogorov equations can be manipulated to find expressions for generating functions and the mean population size. When models become too complex, analytical approaches are no longer feasible. In this case, we will introduce two simulations techniques, the Gillespie and tau-leaping algorithms, that allow us to create numerical realisations of the stochastic process in question. Once a stochastic model has been developed, we may wish to infer the parameters of our model from experimental data. We therefore introduce a Bayesian approach to do this that incorporates our prior beliefs about the parameters, along with the data, to produce a probability distribution for each parameter.

15:30 to 16:00 -- Coffee break
16:00 to 17:00 Jonathan Carruthers Stochastic modelling of processes in Biology I (Tutorial Session 4)

In this workshop, we introduce approaches to study biological populations using stochastic processes. Through deriving the Kolmogorov equations, we can form systems of ODEs that describe the time evolution of the transition probabilities, the probability of the stochastic process being in a particular state at a particular time. For some processes, analytic solutions to these equations can be obtained, however, when this is not possible the Kolmogorov equations can be manipulated to find expressions for generating functions and the mean population size. When models become too complex, analytical approaches are no longer feasible. In this case, we will introduce two simulations techniques, the Gillespie and tau-leaping algorithms, that allow us to create numerical realisations of the stochastic process in question. Once a stochastic model has been developed, we may wish to infer the parameters of our model from experimental data. We therefore introduce a Bayesian approach to do this that incorporates our prior beliefs about the parameters, along with the data, to produce a probability distribution for each parameter.

Thursday, 11 July 2019
Time Speaker Title Resources
09:00 to 10:00 -- Discussions and Collaborations
10:00 to 11:00 Jonathan Carruthers Stochastic modelling of processes in Biology I (Tutorial Session 4)

In this workshop, we introduce approaches to study biological populations using stochastic processes. Through deriving the Kolmogorov equations, we can form systems of ODEs that describe the time evolution of the transition probabilities, the probability of the stochastic process being in a particular state at a particular time. For some processes, analytic solutions to these equations can be obtained, however, when this is not possible the Kolmogorov equations can be manipulated to find expressions for generating functions and the mean population size. When models become too complex, analytical approaches are no longer feasible. In this case, we will introduce two simulations techniques, the Gillespie and tau-leaping algorithms, that allow us to create numerical realisations of the stochastic process in question. Once a stochastic model has been developed, we may wish to infer the parameters of our model from experimental data. We therefore introduce a Bayesian approach to do this that incorporates our prior beliefs about the parameters, along with the data, to produce a probability distribution for each parameter.

11:00 to 11:30 -- Coffee break
11:30 to 12:30 Jonathan Carruthers Stochastic modelling of processes in Biology I (Tutorial Session 4)

In this workshop, we introduce approaches to study biological populations using stochastic processes. Through deriving the Kolmogorov equations, we can form systems of ODEs that describe the time evolution of the transition probabilities, the probability of the stochastic process being in a particular state at a particular time. For some processes, analytic solutions to these equations can be obtained, however, when this is not possible the Kolmogorov equations can be manipulated to find expressions for generating functions and the mean population size. When models become too complex, analytical approaches are no longer feasible. In this case, we will introduce two simulations techniques, the Gillespie and tau-leaping algorithms, that allow us to create numerical realisations of the stochastic process in question. Once a stochastic model has been developed, we may wish to infer the parameters of our model from experimental data. We therefore introduce a Bayesian approach to do this that incorporates our prior beliefs about the parameters, along with the data, to produce a probability distribution for each parameter.

12:30 to 13:30 -- Lunch
13:30 to 14:30 Martín López-García Stochastic modelling of processes in Biology II (Tutorial Session 5)

In this session, we will continue with the mathematical and computational approaches explained in TS4. The focus here will be on how some of the (approximative) numerical results obtained in TS4 by means of Gillespie simulations can be obtained analytically instead, in an exact way, by means of first-step arguments. We will show some applications in the area of mathematical epidemiology, moving from compartmental models for widely homogeneous populations to agent-based oriented approaches for more heterogeneous ones.

14:30 to 15:30 Martín López-García Stochastic modelling of processes in Biology II (Tutorial Session 5)

In this session, we will continue with the mathematical and computational approaches explained in TS4. The focus here will be on how some of the (approximative) numerical results obtained in TS4 by means of Gillespie simulations can be obtained analytically instead, in an exact way, by means of first-step arguments. We will show some applications in the area of mathematical epidemiology, moving from compartmental models for widely homogeneous populations to agent-based oriented approaches for more heterogeneous ones.