Zkoumání sekvenčních parametrů strukturovanosti a nestrukturovanosti proteinů

Sotáková, Patrícia

Much of the current success is driven by complex deep learning methods with high predictive accuracy for protein structure. However, relating these predictions to the un- derlying protein sequence, which ultimately governs both structure and disorder, remains an important challenge. Sequence-based descriptors provide a means to differentiate levels of protein structural organisation. In this work, I propose a conceptually straight- forward approach to analysing protein ensembles using statistical sequence descriptors. Specifically, I investigate whether contextual patterns of amino acid composition can differentiate between ordered and disordered proteins. A computational pipeline was developed to transform protein sequence datasets into trio profiles, where trios (amino acid pairs separated by a defined sequence distance) are classified as over-represented, random, or under-represented, reflecting their enrichment in the dataset. These profiles were then compared and visualised in the context of experimentally determined struc- tural data, better to understand the origins of order- and disorder-related signals. This work demonstrates that the frequency of such sequence-separated amino acid pairs can form profiles that capture aspects of structural context, with the potential to identify signals associated...

guest :: login Digital Repository
		Search		Submit		Help		About