BEGIN:VCALENDAR VERSION:2.0 PRODID:-//DTU.dk//NONSGML DTU.dk//EN CALSCALE:GREGORIAN BEGIN:VEVENT DTSTART:20221124T123000Z DTEND:20221124T153000Z SUMMARY:PhD defense by Alessandro Montemurro DESCRIPTION:
On Thursday 24 November 2022, Alessandro Montemurro will defend his PhD thesis "Improved Immunoinformatic Methods for Rationale T Cell Epitope Discovery".
\n\n
Time: 13:30
\nPlace: Building 341, auditorium 22
\n\n
Supervisor: Professor Morten Nielsen
\nCo-Supervisor: Associate Professor Leon Eyrich Jessen
\n
\n
\n
Assessment committee:
\nAssociate Professor Henrik Nielsen, DTU Health Tech
\nSenior Researcher Peter Meysman, Department of Mathematics and Informatics, University of Antwerp
\nSenior Director, Immunology, Peter Sejr Andersen, ALK
\n
Chairperson:
\nAssociate Professor Elena Papaleo, DTU Health Tech
\n
Abstract:
\nOur immune system is a vast and intricate set of mechanisms specialized to protect the body from the outer world. The role of the immune system cells is to circulate in the body, always screening the surrounding. Specifically, T cells use a receptor on their surface, the T cell receptor (TCR) to scan the peptide-MHC complexes expressed on the surface of the cells. When a T cell encounters a peptide fragment derived from a virus or product of a mutation, an immune response is triggered and a chain of events it activated, aiming to kill the malfunctioning cell. Thus, the TCR-peptide-MHC complex represents the hallmark of T cell-mediated immunity. As the TCRs are highly specific to a pathogen, the immune system evolved so that the TCRs are immensely variable, to ensure the broadest protection possible. This variability is what makes adaptive immunity so powerful, but at the same time, it makes it challenging to study its principles. In the last years, the volume of the TCR specificity data generated is steadily increasing and bioinformatics techniques are needed to analyze this data. Recently, with the advent of single-cell sequencing, the amount of data is growing even faster, and this technology holds the promise of generating large amounts of accurate data. However, this new technology presents new challenges and data-driven approaches to process the data and truly benefit from it.
\nThe aim of this thesis was to advance the current understanding of peptide-MHC recognition by T cell receptors and build machine learning models to predict their interaction. The ability to predict this interaction would make it easier to track the development of infectious diseases and open the door to immunotherapies for cancer or T cell-based vaccine design.
\nWe developed NetTCR-2.0, and an updated version, NetTCR-2.1, two neural network-based frameworks to predict TCR-peptide interactions. We showed that it is possible to successfully predict the binding event between TCRs and peptides, given that the data quality and quantity are appropriate. Lastly, we presented two frameworks, namely ICON and ATRAP, to process the single-cell data and remove artifacts; we showed that NetTCR performs better on the cleaned datasets, compared to the original one, indicating that the two frameworks indeed remove noise from the data, keeping only the signal.
On Thursday 24 November 2022, Alessandro Montemurro will defend his PhD thesis "Improved Immunoinformatic Methods for Rationale T Cell Epitope Discovery".
\n\n
Time: 13:30
\nPlace: Building 341, auditorium 22
\n\n
Supervisor: Professor Morten Nielsen
\nCo-Supervisor: Associate Professor Leon Eyrich Jessen
\n
\n
\n
Assessment committee:
\nAssociate Professor Henrik Nielsen, DTU Health Tech
\nSenior Researcher Peter Meysman, Department of Mathematics and Informatics, University of Antwerp
\nSenior Director, Immunology, Peter Sejr Andersen, ALK
\n
Chairperson:
\nAssociate Professor Elena Papaleo, DTU Health Tech
\n
Abstract:
\nOur immune system is a vast and intricate set of mechanisms specialized to protect the body from the outer world. The role of the immune system cells is to circulate in the body, always screening the surrounding. Specifically, T cells use a receptor on their surface, the T cell receptor (TCR) to scan the peptide-MHC complexes expressed on the surface of the cells. When a T cell encounters a peptide fragment derived from a virus or product of a mutation, an immune response is triggered and a chain of events it activated, aiming to kill the malfunctioning cell. Thus, the TCR-peptide-MHC complex represents the hallmark of T cell-mediated immunity. As the TCRs are highly specific to a pathogen, the immune system evolved so that the TCRs are immensely variable, to ensure the broadest protection possible. This variability is what makes adaptive immunity so powerful, but at the same time, it makes it challenging to study its principles. In the last years, the volume of the TCR specificity data generated is steadily increasing and bioinformatics techniques are needed to analyze this data. Recently, with the advent of single-cell sequencing, the amount of data is growing even faster, and this technology holds the promise of generating large amounts of accurate data. However, this new technology presents new challenges and data-driven approaches to process the data and truly benefit from it.
\nThe aim of this thesis was to advance the current understanding of peptide-MHC recognition by T cell receptors and build machine learning models to predict their interaction. The ability to predict this interaction would make it easier to track the development of infectious diseases and open the door to immunotherapies for cancer or T cell-based vaccine design.
\nWe developed NetTCR-2.0, and an updated version, NetTCR-2.1, two neural network-based frameworks to predict TCR-peptide interactions. We showed that it is possible to successfully predict the binding event between TCRs and peptides, given that the data quality and quantity are appropriate. Lastly, we presented two frameworks, namely ICON and ATRAP, to process the single-cell data and remove artifacts; we showed that NetTCR performs better on the cleaned datasets, compared to the original one, indicating that the two frameworks indeed remove noise from the data, keeping only the signal.