Weighted ensemble oriented method development for increased efficiency of drug design
In drug design, there are many biologically relevant values of interest such as (un)binding rates and free energies. Recent works show that values such as the unbinding rate, koff, may be equally as important to the drug design process as the free energy, ⁸́⁶G. This is due the relationship between koff and the RT for a given drug, as RT can be a predictor of drug efficacy. In order to determine koff, information about the transition path ensemble is required, but the instability and short lifespan of this state makes it difficult to gain atomic level insights with experimental methods. Many computational methods oriented around MD have been developed to predict these values. However, events of interest, such as binding and unbinding events necessary to compute rates and energies, are rare and often occur on prohibitively long timescales ranging from milliseconds to hours. This is problematic as most straight-forward MD simulations are computationally limited to the microsecond timescale. Enhanced sampling algorithms such as the path sampling method WE permit the simulation of these pathways in less computational time, through the resampling process of merging away redundant trajectories in an ensemble, and cloning trajectories of interest. In this thesis, a new resampling algorithm that is based on the JE for use with the WE method is developed that allows the use of short, efficient, nonequilibrium simulations to predict (un)binding free energies with a Lennard-Jones pair. This method is found to be generally inefficient with larger, more protein-ligand like test systems called SAMPL systems (host-guest pairs). In order to better this method, solvent-based features of these systems are studied in detail, and it is determined that high-probability events of interest have very little guest-ion interaction. This is of great interest due to the prevalence of SAMPL systems in method development. It is also found that there is a physical point in the unbinding path for these systems that represents a ''point of no return", or a commitment to unbinding point. With this in mind, a new resampling algorithm based off of the preexisting REVO algorithm is developed. This prevents cloning operations from occurring on trajectories that have surpassed the commitment to unbinding point. This allows the probabilities of these trajectories to remain high when the target state is reached and leaves room open in the ensemble for the exploration of other pathways. This new resampler called ''cutoff-REVO" produces far fewer, but much higher probability unbinding events than its predecessor and more accurately and consistently predicts both ⁸́⁶G and koff for four systems of interest. Overall, this work has provided insights into systems of interest and new means of obtaining information essential to the drug design process.
Read
- In Collections
-
Electronic Theses & Dissertations
- Copyright Status
- In Copyright
- Material Type
-
Theses
- Authors
-
Roussey, Nicole Marie
- Thesis Advisors
-
Dickson, Alex
- Date
- 2022
- Subjects
-
Biochemistry
Drugs--Design
- Program of Study
-
Biochemistry and Molecular Biology - Doctor of Philosophy
- Degree Level
-
Doctoral
- Language
-
English
- Pages
- viii, 117 pages
- ISBN
-
9798352914250
- Permalink
- https://doi.org/doi:10.25335/yfvs-r253