Propostas atribuidas 2024/2025

DEI - FCTUC
Gerado a 2024-12-11 23:50:47 (Europe/Lisbon).
Voltar

Titulo Estágio

Harnessing TCR repertoire to improve immunogenicity prediction

Local do Estágio

Roche Diagnostics GmbH, Penzberg, Germany

Enquadramento

Immunogenic peptides in therapeutic proteins are presented by MHC molecules, which trigger T-cells through the binding of T-cell receptors (TCRs). Current T-cell epitope prediction is based on the likelihood of a peptide being presented by different MHC alleles, even though not all presented peptides would trigger T-cell reactions. On the other hand, existing TCR epitope predictors have been trained on a limited number of peptide epitopes, which perform well only on peptides similar to those in the training set. The disconnection between TCR-pMHC modelling leaves a gap especially in the domain of CD4+ and MHCII interactions. The aim of the project is to improve immunogenicity prediction by integrating information of T-cell receptors, both from the sequence and structure description.

Objetivo

The thesis will involve the curation of a database of peptide epitopes, MHC allele information and TCR sequences. From this database, a model will be built to help reduce false positives predicted from existing MHC presentation predictors. Input featurization will range from sequence, structure and embeddings from deep learning models. A range of machine learning models will be explored for the prediction of TCR-pMHC formation. The resulting model will be used to remove false positives T-cell epitopes to improve correlation to immunogenicity.

Plano de Trabalhos - Semestre 1

Familiarisation with the Roche environment
Review, familiarise and consolidate public data sets
Generate sequence-based embedding
First prototype of machine learning models with different sequence-based combinations
Write-up of interim report

Plano de Trabalhos - Semestre 2

Generate structure-based input and extract structural embedding
Test different ML models, with additional structure-based input combinations
Test and improve correlation of the predicted TCR-reactive epitopes in clinical antibodies and their immunogenicity levels
Evaluate approach on the specific case studies
Document codes and approaches and write-up of master thesis

Condições

The work will be carried out at Roche Diagnostics in Penzberg, Germany. At Roche the student will have at his disposal all the tools and resources (hardware and software) required.

Observações

This work will be developed in Roche Diagnostics GmbH, Penzberg, Germany

Orientador

Catherine Wong
catherine.wong.cw3@roche.com 📩