Molecular Binding Prediction for Drug Design

Summary

Multi-molecular complexes play a key role in the regulation of important biological phenomena. The ability to predict their structure and dynamics is a key challenge in structural biology. This project focuses on the design and implementation of computational models, techniques, and protocols for the prediction of multi-molecular binding, for molecular dynamics simulations, and for the subsequent trajectory analysis. The tools developed are general and have a broad range of applications in the biomedical field and beyond. The core application is currently the study of protein complexes involved in neurodegeneration. Our findings, along with experimental data from our collaborators, at NIH and NIST PML, are contributing to the definition of novel therapeutic strategies for neurodegenerative diseases such as Alzheimer’s.

Description

Computer simulations of large multi-molecular networks are of fundamental importance to elucidate the underlying mechanisms of key biological functions in the human body. Our vision is that only the combination of computational and experimental data can lead to the needed insight into these complex mechanisms at atomic level. This is achieved through a feedback loop between in vitro/in vivo and in silico measurements, which also contributes to the enhancement of both measurement devices and the computational models and tools. Such a feedback loop can be realized only if the timescale of molecular simulations is comparable to the one associated to phenomena of interest as well as to experimental timescales. To this end, efficient computational simulation tools and models need to be developed.

To demonstrate our vision, we are focusing on the pathological complex Cdk5-p25. The kinase Cdk5 is a protein involved in many neuronal functions, whose physiological activator is protein p35. Under neuronal stress, p35 is cleaved into the pathological activator p25, which hyperactivates Cdk5 and leads to the formation of plaques and tangles, hallmarks of neurodegenerative diseases such as Alzheimer’s. In the past a group of NIH-NINDS scientists led by Dr. Harish Pant have discovered peptides, obtained from p25 truncation, that act as selective inhibitors of Cdk5-p25 pathology. To discover more effective drug targets for Cdk5-p25 pathology, we have been focusing on the development of efficient computational models and simulation tools in collaboration with Dr. Sergio Hassan, a computational physicist from the Bioinformatics and Computational Biosciences Branch of NIH-NIAID. Besides the experimental data obtained at NIH-NINDS, measurements of Cdk5-p25 and Cdk5-p35 activity and peptide-based inhibition are carried out at NIST-PML by a group of scientists led by Dr. Arvind Balijepalli who designed a novel biosensor, capable of measuring Ph change in a solution at high resolution. Biosensor data has a fast turnaround, and it will be used to validate the computer models and tools.

This project operates in the computer-aided drug design domain. Inhibition mechanisms are studied using computer simulations and tested in vitro, providing the envisioned feedback loop between in vitro/in vivo and in silico measurements. The developed computational models and tools are general and can be used to study other multi-molecular networks of interest for the biomedical field and beyond.

A list of goals associated with the project is listed as follows:

• Develop efficient computational models and tools to study multi-molecular networks at atomic level and at timescales comparable to the ones from experimental data/evidence.
• Develop analysis tools to identify and quantitively characterize relevant dynamics from molecular trajectories at atomic level.
• Define validation procedures for the computational models and tools based on available measurement data.
• Use the developed computational models and tools to study Cdk5-p25 pathological complex and its peptide-based inhibition.
• Validate the developed computational models and tools using data and measurements obtained from the Cdk5-p25 pathological complex and its peptide-based inhibition.

Major Accomplishments

• We developed multi-resolution computer models for efficient molecular binding prediction, using coarse-grained and all-atom representations.
• We developed efficient multi-stage molecular binding prediction techniques. For the early stages of complexation, we used simplified force fields and fast Monte Carlo sampling procedures. Then, progressively more refined models based on implicit solvent models and all-atom solvent representation were used.
• We defined all-atom molecular dynamics simulation protocols for the refinement of the predicted binding modes and for the study of the discovered binding modes.
• We developed targeted analysis techniques to identify the prevailing dynamics from the obtained molecular trajectories, and to identify the specific amino acid groups and interaction patterns involved in the biological phenomena of interest (e.g., pharmacophore).
• We applied the above computational tools to the study of p5-based inhibition of AD Cdk5-p25 pathology. Potentially inhibitory Cdk5-p5 binding modes were identified, and the pharmacophore for one mode was characterized, consistent with competitive inhibition.
• We contributed to the definition of a biosensor-based measurement assay targeting the activation and p5-based inhibition of the AD Cdk5-p25 pathological complex. Measurement data will be used to validate the developed computational models and tools.