This thesis focused on modeling of ligand-protein binding with computational methods based on molecular dynamics. Understanding this process is crucial for the design and discovery of new drugs and the use of computational methods to support experimental research in this field is constantly growing. Nowadays, thanks to the increasing computer power, it is possible to study the complete ligand binding/unbinding process and obtain estimate on thermodynamic and kinetic properties. In view of this, during my PhD, different advanced classical molecular dynamics (MD) methods were employed and compared to identify an effective computational approach for studying ligand binding/unbinding processes. Specifically, a protocol based on combination of the steered MD (sMD) and the Metadynamics (MetaD) with Path Collective Variables (PCVs) approaches was developed with the aim of using the advantages of both methods to obtain a complete description of the process. While the sMD method was employed to investigate different unbinding pathways and identify the preferred one, MetaD with PCVs was used to determine more accurately the binding free energy. The proposed protocol was successfully applied to study ligand binding to the Hypoxia Inducible Factor (HIF-2α) and it demonstrated to be effective for simulations performed both on a known ligand-protein X-ray structure and on a docking pose. On the other hand, most of the MD methods requires the production of several replicas or long simulations to sample the binding/unbinding event several times in order to obtain a reliable statistics of the process. This produces the need of methods able to analyze all the simulated events at once and to provide a clearly interpretable picture of the differences in the sampled pathways. For this reason, a tool based on the self-organizing maps (SOMs) was developed. The PathDetect-SOM (Pathway Detection on SOM) tool, exploiting the advantages of the topological ordering of the SOM, allowing to visually represent the binding paths sampled during different MD events/replicas in a clear bidimensional representation. In addition, hints on the kinetic and thermodynamic properties of the process can be derived. The tool was successfully applied to different study-cases to demonstrate its general applicability. Furthermore, as part of a project performed at the Jülich research center (Institute of Advanced Simulations and Institute for Neuroscience and Medicine) under the supervision of Prof. Paolo Carloni, a novel hybrid quantum mechanics/molecular mechanics (QM/MM) interface (MiMiC) was tested. The code, that allows QM/MM molecular dynamics simulations of biomolecular systems, was applied to the mitogen-activated protein kinase p38 in complex with the 2g ligand to investigate the ligand unbinding process. The focus was on the first step of the process involving the dynamics of the ligand in its bound state. QM/MM MD simulations were effective in describing ligand-protein interactions accurately. In particular, by monitoring the change of the atomic charges during the simulation and calculating the electronic density difference between the ligand in its bound state and in vacuum, insights into the polarization effects of the protein electric field onto the ligand were obtained. It is expected that these effects, albeit small in the bound state, become very important in the following steps of the unbinding process.

Questa tesi è incentrata sulla modellazione del binding ligando-proteina con metodi computazionali basati sulla dinamica molecolare. La comprensione di questo processo è fondamentale per la progettazione e la scoperta di nuovi farmaci e l'uso di metodi computazionali per supportare la ricerca sperimentale in questo campo è in costante crescita. Oggi, grazie alla crescente potenza dei computer, è possibile studiare l'intero processo di binding/unbinding del ligando e ottenere stime sulle proprietà termodinamiche e cinetiche. Alla luce di ciò, durante il mio dottorato, diversi metodi avanzati di dinamica molecolare classica (MD) sono stati impiegati e confrontati per identificare un approccio computazionale efficace per studiare i processi di binding/unbinding dei ligandi. In particolare, è stato sviluppato un protocollo basato sulla combinazione degli approcci steered MD (sMD) e Metadinamica (MetaD) con Path Collective Variables (PCVs) con lo scopo di utilizzare i vantaggi di entrambi i metodi per ottenere una descrizione completa del processo. Mentre il metodo SMD è stato impiegato per studiare diversi percorsi di disassociazione e identificare quello preferito, la MetaD con le PCVs è stato utilizzato per determinare più accuratamente l'energia libera di legame. Il protocollo proposto è stato applicato con successo per studiare il legame del ligando al fattore inducibile dell'ipossia (HIF-2α) e ha dimostrato di essere efficace per le simulazioni effettuate sia su una struttura a raggi X nota del ligando-proteina che su una posa di docking. D'altra parte, la maggior parte dei metodi MD richiede la produzione di diverse repliche o lunghe simulazioni per campionare più volte l'evento di binding/unbinding al fine di ottenere una statistica affidabile del processo. Questo produce la necessità di metodi in grado di analizzare tutti gli eventi simulati in una sola volta e di fornire un quadro chiaramente interpretabile delle differenze nei pathway campionati. Per questo motivo, è stato sviluppato un tool basato sulle mappe auto-organizzanti (SOM). Lo strumento PathDetect-SOM (Pathway Detection on SOM), sfruttando i vantaggi dell'ordinamento topologico della SOM, permette di rappresentare visivamente i percorsi di legame campionati durante diversi eventi/repliche MD in una chiara rappresentazione bidimensionale. Inoltre, possono essere derivati suggerimenti sulle proprietà cinetiche e termodinamiche del processo. Lo strumento è stato applicato con successo a diversi casi di studio per dimostrare la sua applicabilità generale. Inoltre, come parte di un progetto eseguito presso il centro di ricerca Jülich (Istituto di Simulazioni Avanzate e Istituto di Neuroscienze e Medicina) sotto la supervisione del Prof. Paolo Carloni, è stata testata una nuova interfaccia ibrida di meccanica quantistica/meccanica molecolare (QM/MM) (MiMiC). Il codice, che permette simulazioni di dinamica molecolare QM/MM di sistemi biomolecolari, è stato applicato alla proteina chinasi mitogeno-attivata p38 in complesso con il ligando 2g per studiare il processo di unbinding del ligando. L'attenzione si è concentrata sulla prima fase del processo che coinvolge la dinamica del ligando nel suo stato legato. Le simulazioni MD QM/MM sono state efficaci nel descrivere accuratamente le interazioni ligando-proteina. In particolare, monitorando il cambiamento delle cariche atomiche durante la simulazione e calcolando la differenza di densità elettronica tra il ligando nel suo stato legato e nel vuoto, sono stati ottenuti approfondimenti sugli effetti di polarizzazione del campo elettrico della proteina sul ligando. Ci si aspetta che questi effetti, anche se piccoli nello stato legato, diventino molto importanti nelle fasi successive del processo di unbinding.

(2022). MODELING OF LIGAND-PROTEIN BINDING WITH ADVANCED MOLECULAR DYNAMICS METHODS. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2022).

MODELING OF LIGAND-PROTEIN BINDING WITH ADVANCED MOLECULAR DYNAMICS METHODS

CALLEA, LARA
2022

Abstract

This thesis focused on modeling of ligand-protein binding with computational methods based on molecular dynamics. Understanding this process is crucial for the design and discovery of new drugs and the use of computational methods to support experimental research in this field is constantly growing. Nowadays, thanks to the increasing computer power, it is possible to study the complete ligand binding/unbinding process and obtain estimate on thermodynamic and kinetic properties. In view of this, during my PhD, different advanced classical molecular dynamics (MD) methods were employed and compared to identify an effective computational approach for studying ligand binding/unbinding processes. Specifically, a protocol based on combination of the steered MD (sMD) and the Metadynamics (MetaD) with Path Collective Variables (PCVs) approaches was developed with the aim of using the advantages of both methods to obtain a complete description of the process. While the sMD method was employed to investigate different unbinding pathways and identify the preferred one, MetaD with PCVs was used to determine more accurately the binding free energy. The proposed protocol was successfully applied to study ligand binding to the Hypoxia Inducible Factor (HIF-2α) and it demonstrated to be effective for simulations performed both on a known ligand-protein X-ray structure and on a docking pose. On the other hand, most of the MD methods requires the production of several replicas or long simulations to sample the binding/unbinding event several times in order to obtain a reliable statistics of the process. This produces the need of methods able to analyze all the simulated events at once and to provide a clearly interpretable picture of the differences in the sampled pathways. For this reason, a tool based on the self-organizing maps (SOMs) was developed. The PathDetect-SOM (Pathway Detection on SOM) tool, exploiting the advantages of the topological ordering of the SOM, allowing to visually represent the binding paths sampled during different MD events/replicas in a clear bidimensional representation. In addition, hints on the kinetic and thermodynamic properties of the process can be derived. The tool was successfully applied to different study-cases to demonstrate its general applicability. Furthermore, as part of a project performed at the Jülich research center (Institute of Advanced Simulations and Institute for Neuroscience and Medicine) under the supervision of Prof. Paolo Carloni, a novel hybrid quantum mechanics/molecular mechanics (QM/MM) interface (MiMiC) was tested. The code, that allows QM/MM molecular dynamics simulations of biomolecular systems, was applied to the mitogen-activated protein kinase p38 in complex with the 2g ligand to investigate the ligand unbinding process. The focus was on the first step of the process involving the dynamics of the ligand in its bound state. QM/MM MD simulations were effective in describing ligand-protein interactions accurately. In particular, by monitoring the change of the atomic charges during the simulation and calculating the electronic density difference between the ligand in its bound state and in vacuum, insights into the polarization effects of the protein electric field onto the ligand were obtained. It is expected that these effects, albeit small in the bound state, become very important in the following steps of the unbinding process.
BONATI, LAURA
GRECO, CLAUDIO
binding ligandi; interazioni proteina; dinamica molecolare; SOM; MiMiC
ligand binding; protein interactions; molecular dynamics; SOM; MiMiC
CHIM/02 - CHIMICA FISICA
English
2-mag-2022
SCIENZE CHIMICHE, GEOLOGICHE E AMBIENTALI
34
2020/2021
embargoed_20240502
(2022). MODELING OF LIGAND-PROTEIN BINDING WITH ADVANCED MOLECULAR DYNAMICS METHODS. (Tesi di dottorato, Università degli Studi di Milano-Bicocca, 2022).
File in questo prodotto:
File Dimensione Formato  
phd_unimib_748402.pdf

embargo fino al 02/05/2024

Descrizione: Tesi
Tipologia di allegato: Doctoral thesis
Dimensione 48.23 MB
Formato Adobe PDF
48.23 MB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/10281/374733
Citazioni
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
Social impact