ABSTRACT

Virtual screening (VS) of compounds for possible drug leads requires identifying the relatively few candidates, out of perhaps many thousands, which can bind with significant affinity (typically 100

µ

M or better) to a target of a known structure. An enrichment factor (EF) can be defined as [1]

(9.1)

where

a

is the number of active compounds in the

n

top-ranked compounds of a total database of

N

compounds of which

A

are active. Successful screening implies EF >> 1. This in turn requires the identification of the best ligand conformation/position/orientation (pose) in the target binding site, that is, the solution of the docking problem. Solution of the docking problem in turn requires the ability to accurately calculate the binding affinity of a given pose (at least relative to another pose), which is the solution of the binding problem. Clearly VS is a formidable task that must ultimately depend on the solution of the binding problem at some level of accuracy. The binding affinity defined here as the association constant K

= [PL]/[P][L] (or often in the literature by the dissociation constant K

= 1/K

), or equivalently, the binding free energy

G

= -kTln(K

), is an equilibrium thermodynamic property that is connected to the underlying molecular and atomic detail structure of the target, ligand, and solvent through the laws of statistical mechanics. The view taken in this chapter is that any method that successfully scores or ranks candidates or

EF a n

A N =

/ /

poses must be doing it by emulating, however approximately, the true binding free energy. This must be so whether one calls the ranking method a scoring function, a filter function, a screening potential, and even if it was derived with the explicit lesser goal of “not trying to calculate a real binding free energy.” As they say, “if it walks like a duck and talks like a duck....” Referring to the enrichment factor, Equation 9.1, even if EF >> 1 is achieved in an initial screen simply by dropping out a large number of compounds that are difficult or impossible to fit into the binding site (i.e., a shape filter type function), the criteria used to do this must emulate, in a crude way and over the whole range of compounds, the difference in actual binding free energy between the worst compounds and the others.