Computational approaches for the prioritisation of functional phosphorylation sites in yeast

Protein phosphorylation is the main post translational modification that is used by cells during signal transduction. Kinases catalyse the phosphorylation of proteins at predominantly serine, threonine and tyrosine amino acids in response to specific signals, leading to a relay of interactions, new phosphorylations and protein translocations e.g. to the plasma membrane or the nucleus, that together orchestrate the response of cells to their environment in a highly dynamic and regulated manner. Using mass spectrometry-based phosphoproteomics methods we are now able to get snapshots of the ‘signalling’ state of the cells, i.e. signatures of quantified phosphosites in given conditions and time points under study. Using such technologies thousands of phosphosites have been discovered and measured in various conditions, opening the door to an improved understanding of the regulation of signalling responses through phosphorylation. However, only 5% of these measured phosphosites are annotated with an upstream kinase or a functional implication. Moreover, it is now well-understood that not all phosphosites measured are functional, with a fraction likely being either biological or technical noise. To be able to understand the phospho-regulation of signalling proteins it is critical to first identify which phosphosites are functional and then prioritise them for functional studies to place them within the context of current knowledge. During this PhD I built upon a recent study that used machine learning to assign a functional score to 100,000 phosphosites from human cells. Specifically, I focused on yeast as a model organism, as it is a well-studied, tractable and easy-to-manipulate system that has been proven pivotal in understanding fundamental biological principles generally but also often translatable to human biology. In Chapter 1 I present an introduction to the field, highlighting the importance of the question, existing approaches to tackle it, the main gaps, and briefly mentioning how I addressed one of these gaps in this thesis. In Chapter 2 I present a machine learning model that used 12 features related to functional regulation, evolution, and protein structure, that proved important for predicting a functional score for yeast phosphosites. In Chapter 3 I present two case studies demonstrating how prioritising functional phosphosites can help design downstream experiments to assign function to yeast phosphosites. Specifically, I identify modules associated with specific kinase phosphosites or protein-protein interaction interfaces and use networks to generate hypotheses about non-annotated phosphosites. Chapters 4 and 5 explore the use of additional features for the model, namely the relationship of phosphorylation sites with protein localisation signals and the identification of phosphorylated ‘hotspots’ on domain-adjacent regions of proteins. In the process I discovered that phosphorylation sites prefer to be further away from N-terminal localisation signals and found several phosphosites that consistently appear between or before/after specific domains, with implications for their function. Overall, the results of this thesis present a valuable resource for the yeast community and beyond, allowing the prioritisation of the yeast phosphoproteome for functional studies. In the longer term, I expect this work to lead to an improved understanding of the principles underpinning signalling regulation and providing a framework to study the function of the ‘dark’ signalling space in yeast.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2024

Author dc:creator

• Petsalaki, Eirini

Advisor dc:contributor.advisor

• Cortés-Ciriano, Isidro

• functional phosphorylation sites
• yeast

Licence

• https://www.repository.cam.ac.uk/bitstreams/be3d86b5-3e82-4feb-8d6d-ecd82405d8ba/download
• https://creativecommons.org/licenses/by/4.0/

Language dc:language

eng

Author Identifier

0000-0002-7514-4362

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/399427