Use this URL to cite or link to this record in EThOS:
Title: Computational investigations of backbone dynamics in intrinsically disordered proteins
Author: Kosciolek, T. P.
ISNI:       0000 0004 8503 2756
Awarding Body: UCL (University College London)
Current Institution: University College London (University of London)
Date of Award: 2016
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Intrinsically disordered proteins (IDPs), due to their dynamic nature, play important roles in molecular recognition, signalling, regulation, or binding of nucleic acids. IDPs have been extensively studied computationally in terms of binary disorder/order classification. This approach has proven to be fruitful and enabled researchers to estimate the amount of disorder in prokaryotic and eukaryotic genomes. Other computational methods - molecular dynamics, or other simulation techniques, require a starting structure. However, there are no approaches permitting insight into the behaviour of disordered ensembles from sequence alone. Such a method would facilitate the study of proteins of unknown structures, help to obtain a better classification of the disordered regions, and the design disorder-to-order transitions. In this work, I develop FRAGFOLD-IDP, a method to address this issue. Using a fragment-based structure prediction approach - FRAGFOLD, I generate the ensembles of IDPs and show that the features extracted from them correspond well with the backbone dynamics of NMR ensembles deposited in the PDB. FRAGFOLD-IDP predictions significantly improve over a naïve approach and help to get a better insight into the dynamics of the disordered ensembles. The results also show it is not necessary to predict the correct fold of the protein to reliably assign per-residue fluctuations to the sequence in question. This suggests that disorder is a local property and it does not depend on the protein fold. Next, I validate FRAGFOLD-IDP on the disorder classification task and show that the method performs comparably to machine learning-based approaches designed specifically for this task. I also found that FRAGFOLD-IDP produces results on par with DynaMine, a machine learning approach to predict the NMR order parameters and that the results of both methods are not correlated. Thus, I constructed a consensus neural network predictor, which takes the results of FRAGFOLD-IDP, DynaMine and physicochemical features to predict per-residue fluctuations, improving upon both input methods.
Supervisor: Jones, D. T. Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available