التفاصيل البيبلوغرافية
| العنوان: |
Computational methods that predict residence times of GPCR ligands |
| المؤلفون: |
Potterton, Andrew Graven |
| بيانات النشر: |
University College London (University of London), 2020. |
| سنة النشر: |
2020 |
| المجموعة: |
University College London (University of London) |
| مصطلحات موضوعية: |
612 |
| الوصف: |
This thesis describes the development and analysis of different computational meth- ods that predict drug-target residence time, the duration of ligand binding at its target. Residence time has been shown to be a better surrogate of in vivo efficacy than equilib- rium binding affinity. All methods were applied to ligands acting at G protein-coupled receptors (GPCRs). GPCRs are a massive pharmaceutical target, with approximately one third of approved drugs acting at a GPCR. Three sets of computational methods were used to predict residence time. The first set are machine learning (ML) approaches trained only on ligand descriptors, the second set are molecular dynamics (MD) approaches and the final set of methods combine ML and MD. All three sets of methods were developed against a database of GPCR ligand kinetic data compiled from published sources. Different ML approaches were applied to ligands of 20 GPCRs. Both the principle component analysis and the multi-linear regression revealed properties relating to size of the ligand have a correlation to residence time. The first of two MD-based approaches was a steered MD method. By applying this method to 17 A2A receptor ligands, it was found that the changing interaction energies made by the dissociating ligand to both the receptor residues and to water correlated strongly with residence time values. The second MD-approach is a recently- published method, τ-RAMD. Results from τ-RAMD were found to correlate more strongly with molecular weight than residence time. Finally, ML models were trained on ensembles of short MD simulations of 259 GPCR-ligand complexes. The model with the highest accuracy was a gradient boosted regressor model trained on a combination of ligand molecular descriptors and the non- bonded interaction energy between the receptor-bound ligands and water. |
| نوع الوثيقة: |
Electronic Thesis or Dissertation |
| اللغة: |
English |
| URL الوصول: |
https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.822775 |
| رقم الانضمام: |
edsble.822775 |
| قاعدة البيانات: |
British Library EThOS |