Fachbereich Informatik - Aktuell

07.04.2017 10:46

Disputation Benjamin Schubert

am Dienstag, 25. April 2017 um 12.45 Uhr in Raum A104, Sand 1.

Advanced Immunoinformatics Approaches for Precision Medicine

Berichterstatter 1: Prof. Oliver Kohlbacher
Berichterstatter 2: Prof. Can Kesmir


Genomic sequencing and other '-omic' technologies are slowly changing biomedical practice. As a result, patients now can be treated based on their molecular profile. Especially the immune system's variability, in particular that of the human leukocyte antigen (HLA) gene cluster, makes such a paradigm indispensable when treating illnesses such as cancer, autoimmune diseases, or infectious diseases. It can be, however, costly and time- consuming to determine the HLA genotype with traditional means, as these methods do not utilize often pre-existing sequencing data. We therefore proposed an algorithmic approach that can use these data sources to infer the HLA genotype. HLA genotyping inference can be cast into a set covering problem under special biological constraints and can be solved efficiently via integer linear programming. Our proposed approach outperformed previously published methods and remains one of the most accurate methods to date.
We then introduced two applications in which a HLA-based stratification is vital for the efficacy of the treatment and the reduction of its adverse effects. In the first example, we dealt with the optimal design of string-of-beads vaccines (SOB). We developed a mathematical model that maximizes the efficacy of such vaccines while minimizing their side effects based on a given HLA distribution. Comparisons of our optimally designed SOB with experimentally tested designs yielded promising results. In the second example, we considered the problem of anti-drug antibody (ADA) formation of biotherapeutics caused by HLA presented peptides. We combined a new statistical model for mutation effect prediction together with a quantitative measure of immunogenicity to formulate an optimization problem that finds alterations to reduce the risk of ADA formation. To efficiently solve this bi-objective problem, we developed a distributed solver that is up to 25-times faster than state-of-the art solvers. We used our approach to design the C2 domain of factor VIII, which is linked to ADA formation in hemophilia A. Our experimental evaluations of the proposed designs are encouraging and demonstrate the prospects of our approach.
Bioinformatics is an integral part of modern biomedical research. The translation of advanced methods into clinical use is often complicated. To ease the translation, we developed a programming library for computational immunology and used it to implement a Galaxy-based web server for vaccine design and a KNIME extension for desktop PCs. These platforms allow researchers to develop their own immunoinformatics workflows utilizing the platform's graphical programming capabilities.