Testing, understanding, and automating applications of alchemical free energy calculations
- Author(s): Klimovich, Pavel
- Advisor(s): Mobley, David
- et al.
Free energy calculations based on molecular dynamics simulations show considerable promise for applications ranging from drug discovery to prediction of physical properties and structure-function studies.
Currently, the most important challenges in this field are primarily in two areas. One is a direct corollary of the finite simulation length and involves problems of obtaining reliable free energy estimates when sampling is inadequate. The other one can be classified as a logistical challenge -- that of obtaining the desired results via as automated a process as possible. This includes such problems as planning of free energy calculations, and their set up and analysis. Here, I present several studies in which I address these challenges.
Chapter 1 focuses on the importance of accounting for ligand binding mode interconversions in relative binding free energy calculations, emphasizing lengthening simulations does not simply eliminate all sampling problem. In Chapter 2 -- a description of the work done as a result of participating in the SAMPL09 contest -- we show that within a given force field framework the conformational changes of
even small, drug-like, molecules can be quite slow, forcing us to employ biased sampling techniques to obtain correct transfer free energies. In addition, the work helped reveal deficiencies of current force fields in the description of hydroxyl-containing molecules.
An attempt to address logistical challenges resulted in the creating of two pieces of software. Chapter 3 describes a tool that automates setup of relative free energy calculations.
Chapter 4 reviews best practices for analysis of alchemical free energy calculations and introduces alchemical-analysis.py, a freely available software that implements the analysis
practices for several simulation packages and features a variety of innovative statistical and graphical ways of assessing the quality of the analyzed data.