Individualised infection medicine
Host-pathogen interactions are highly diverse between different individuals. The outcome of an infection with the same pathogen can be mild of fatal depending of particular properties of the infected individual at the time of infection. Time-resolved measurements of the right biomarkers reflect a specific state of the disease and the host’s immune response. Data taken early after disease onset already contain the information on disease progression. We use mathematical models to stratify patients into groups and to develop treatment strategies adapted to the respective patient groups. This work supports the overarching goal to design specific therapeutic strategies tailored to the individual patient.
Together with Georg Pongratz (Düsseldorf, Germany), we developed a mathematical model of the cytokine response to Ebola virus. By calibrating the model based on datasets from different patients, we were able to predict key differences in the immune response between survivors, fatalities and asymptomatic infections. In particular, we found that differential capability of virus to induce TNF-α (representing innate immune response in the model) is the key mechanism that can explain the cytokine response data of these patients, while differential properties of the virus, or differential induction of IFN-γ (representing cytotoxic T cell response) were not decisive.
Model predictions can be used to stratify high-risk infected patients based on serum cytokine levels upon initial presentation to the clinics. The model predicts that an immune-interfering therapy, which is characterized by improving viral control by exogenous IFN-γ and inhibiting excessive inflammation by IL-10, has the potential to rescue patients classified as fatal Ebola virus infections. Future work will be focused on using cytokine measurements from individual patients to optimize the dose and timing of the proposed therapy.
Despite improvements in vaccination, influenza remains one of the most common diseases and a significant threat to public health, and effective prevention remains challenging due to the limited time window to prevent the spread of new viral strains. As seen in the most recent outbreaks, influenza has a high potential for pandemic spread and health care providers as well as pharmaceutical companies have difficulties coping with the peak demand generated by sudden outbreaks. A successful strategy for reacting to such outbreaks currently includes not only vaccination, but also antiviral treatment. The most effective known antiviral, oseltamivir, inhibits a viral protein called neuraminidase and is able to reduce symptoms and spreading of the disease.
The current treatment schedule includes two daily fixed doses of oseltamivir. While this has proven successful in clinical settings, the limited capacity for stockpiling oseltamivir leads to shortness in supply of the drug in case of an influenza outbreak, prompting the question whether a more efficient use of this limited and expensive resource is possible.
In this project, we employ an approach from control engineering in order to find the optimal dose necessary to treat individual patients. In this context, it is important to note that the drug acts by limiting the spread of the virus within the body, allowing the immune system more room for clearing the virus. Hence, the response to oseltamivir is highly individual and depends on the patient’s immune status, co-morbidities, and general health. Employing a method from control engineering known as adaptive feedback control, we predict the optimal dose and treatment schedule for an individual patient based on measurements of immune effector cells and viral load. Preliminary results show that with this adaptive control-based drug scheduling, the same reduction in viral load can be achieved by using approximately half of the amount of oseltamivir. These promising results warrant further investigation towards a clinically applicable tool for automatically devising a treatment protocol for influenza patients on an individual basis.
Influenza is one of the most common viral diseases and leads to significant morbidity, mortality, and loss of productivity each year. While vaccination against influenza works fine for the general population, it does not work equally well in every patient. Unfortunately, of all patient groups, those with the highest risk are at the same time those that show the least successful vaccinations. With the threat that influenza poses for the individual patient as well as public health, improving vaccination success has a high priority in influenza research. The key objective in this project is to understand the differences in immune parameters of patients with a good and those with a poor vaccination response and to design vaccines tailored to individual patients in order to restore the vaccination success in poor responders.
The development of an efficient antibody response depends on an evolutionary process called affinity maturation in which B cells, the primary producers of specific antibodies, undergo multiple rounds of mutation, selection, and division. Those B cells with higher affinities to the influenza virus become gradually more dominant in this process, leading to the production of highly specific antibody in the end. This process is key to successful vaccination: upon administration of the vaccine, the affinity maturation process in specialised structures called germinal centres is initiated and produces B cells specific for the vaccine, which protect against future encounters of the real virus.
In this project, we employ stochastic and agent-based mathematical models to simulate a) the initiation of a germinal centre reaction, and b) the affinity maturation process in the course of vaccinations. To identify key differences between a successful and an unsuccessful vaccination, we combine extensive clinical data with our theoretical approach. A variety of immune parameters is measured in patients before vaccination and associations of these biomarkers with vaccination success are identified using machine learning techniques. Key immune parameters are included in the mathematical models, allowing for optimised vaccination specifically in those patients with a poor vaccination response. The overarching goal is to develop a tool for clinical use that can predict the optimal vaccination strategy according to easily measured immune parameters of individuals.
Tas JMJ, Mesin L, Pasqual G, Targ S, Jacobsen JT, Mano YM, Chen CS, Weill JC, Reynaud CA, Browne EP, Meyer-Hermann M, Victora GD. Visualizing antibody affinity maturation in germinal centers. Science 351 (2016) 1048-1054.
- Gang Zhao
- Ghazal Montaseri
- Sebastian Binder
- Philippe A. Robert
- Michael Meyer-Hermann
- Systems Immunology- Prof. Dr. Michael Meyer-Hermann