Student Projects

Title: The role and effects of partitioning

Aim: The aim of this project is to identify which partitioning method is the most meaningful for data redundancy and least computationally expensive to obtain the best possible predictions when using Deep Learning.

The project: The student will be working with different partitioning methods to study their effect and role in Deep Learning predictions.

Title: Benchmarking neoepitope prediction methods

Aim: Study and compare state of the art neoepitope prediction methods and their relevance in current immunotherapy strategies

Title: Deep Learning for CDR3 structural feature predictions

Aim: Improve structural feature predictions of highly variable CDR3 regions using Deep Learning

Title: The effect of checkpoint inhibitor therapy on lymphocyte repertoire composition

Aim: The project aim is to investigate available repertoire sequencing data before and after checkpoint inhibitor therapy, and using computational tools to predict the immunogenicity of sequences

Title: Lyra 2.0: furtherance of existing template based prediction

Aim: Optimise Lyra for improved predictions of protein structure models

Title: Developing a similarity metric for complementarity determining region 3 (CDR3) of T cell receptor ß (TCR ß) based on 3D conformational predictions

Aim: The goal of the project is to produce a metric to calculate the similarity between CDR3 regions of the TCRβ from primary sequences, however based on 3D structural predictions.

Title: Diversity of TCR repertoires: state of the art and prospects

Aim: Exploring current immune repertoire diversity analysis and their implications in patient outcome prediction and therapy engineering

Title: Machine Learning for pairing structural data

Aim: Majority of high-throughput sequencing data available in public repositories is unpaired and not on single-cell level. The aim of the project is applying machine learning algorithms in an attempt to match the quality of existing data with those obtained with more recent techniques, thus improving the informational yield

Title: Embeddings for improved immune system relevant predictions

Aim: Studying multiple embedding methods for context learning, which will improve Deep Learning prediction methods for immune system relevant problems

Title: Pysam package optimisation

Aim: Reducing computational cost and data scaling for the Pysam module in relation to current projects