| dc.description.abstract | Diseases are multi-scale processes, with the precise combination of events at different scales both qualitatively and quantitatively dictating the final phenotype. Recognising that immune response is a key determinant of this process, we propose novel algorithms to study several aspects of a disease including susceptibility of an individual, spread of disease in a population, and vaccine efficacy. For each tool developed, we carry out a validation exercise by testing in cases with known expected output, and then apply it to gain new insights.
An adaptive T-cell immune response depends on successful presentation of peptide antigens (epitopes), in complex with the human leukocyte antigen (HLA). Our core hypothesis is that susceptibility to a disease, such as H1N1 influenza, is inversely proportional to the number of high affinity viral epitopes an individual's HLA genotype can present. We classify individuals into sub-populations according to their level of susceptibility, and incorporate this into standard epidemic modelling. To the best of my knowledge, this was the first method to incorporate real immune genetic data into epidemic models. This work resulted in the insight that larger genetic diversity at the level of immune response, leading to the presence of sub-populations with a broad distribution of susceptibilities, protects against the spread of influenza in a population. We also identify correlations between population-scale parameters affected by genetic heterogeneity and the final epidemic size.
The ability to present high affinity peptides also dictates how individuals respond to a vaccine. We use this to aid in vaccine design, and propose OptiNeo, an algorithm to pick a minimal set of antigenic peptides ensuring maximal population coverage. OptiNeo associates peptides with individuals (represented by their HLA genotype) if at least 1 of their HLA alleles can bind with high affinity to that peptide. It then uses set cover to identify a minimal set of peptides to cover the population. This approach uses the commonalities in immune response among individuals to design peptide ensemble vaccine candidates which can be expected to have high efficacy in the largest fraction of the population, only missing individuals whose HLA genotype cannot present any of the considered peptides with high affinity. We apply OptiNeo to design a peptide vaccine against tuberculosis.
We also propose two methods, PathExt and EpiTracer, to facilitate unbiased systems modelling. PathExt identifies differentially active paths when a control is available and most active paths otherwise, in an omics-integrated biological network. EpiTracer interrogates these paths to identify key players involved in spreading a perturbation or responding to it. These methods can extract characteristic genes and pathways even when only a single sample is available, and in the absence of an appropriate control. We apply these to identify a common core in the response of M.tb to drug exposure. This enables a better understanding of the pathways which might be involved in drug resistance, and provides a starting point for intervention.
The work presented in this thesis contributes novel methods to the toolbox for disease modelling, to understand immune response, integrate genetic information into epidemic models, design vaccines with general applicability, and gain mechanistic insights into a system. All the methods developed here are general in nature, and can be used to study any infectious disease or tumour. Each method has been independently validated, and has also been applied to gain new insights. | en_US |
