Electronic Thesis and Dissertation Repository

Thesis Format

Monograph

Degree

Master of Science

Program

Computer Science

Supervisor

Ilie, Lucian

Abstract

This thesis investigates the application of Explainable Artificial Intelligence (XAI) techniques to deep learning models in the field of protein analysis, specifically targeting protein language models and protein interaction site prediction models. Despite the increasing adoption of these sophisticated deep learning models in bioinformatics, their intrinsic complexity often results in a black-box nature, obscuring the understanding of their decision-making processes.

This research represents a thorough effort to integrate explanation methods within this context. We analyze the resulting interpretations using biological-specific statistical tests to enhance the transparency and interpretability of the models. This work evaluates the efficacy of current XAI methods applied to protein analysis through a comprehensive set of experiments.

Summary for Lay Audience

Deep learning models have demonstrated impressive success across various fields, and as their performance improves in different applications, their architecture becomes more complex. Despite their marvellous performance, they are often referred to as "black boxes" due to their opaque decision-making processes. As their complexity increases, their interpretability tends to decrease even as their performance increases. This lack of transparency becomes particularly critical in fields such as bioinformatics, where the consequences of their decision are significant.

This necessity drives our research into interpretation methods specifically for proteins. We have adapted existing interpretation techniques to our protein language models and protein interaction site prediction problems to understand better what the model learned. This work aims to ensure that the deep learning models used in bioinformatics are not only robust but also comprehensible and trustworthy, helping experts make better decisions in their research. This thesis presents a detailed examination of how effective current explainable artificial intelligence methods are in making these sophisticated models more transparent and interpretable.

Download

Share

COinS