A major bottleneck in computational drug-design is the accurate computation of the binding affinity of a ligand-macromolecule interaction from first principles. When a series of analogous ligands with known affinities become available, quantitative structure-activity relationship (QSAR) development offers a viable alternative for binding affinity prediction. There, through statistical analysis of the structure-activity data-set, key physico-chemical characteristics of the ligands can be identified and their relative importance quantified, which can then be used for in silico screening of the next generation of drug leads.

However, the quantification of QSAR descriptors is ambiguous and often conformation dependent, and there is no universal descriptor set that can be applied to every system. There is therefore ample scope for the development of new QSAR descriptors to describe ligand-macromolecule interactions. We present here two new QSAR descriptors based on the interaction forces at the ligand-macromolecule binding interface, which together with the atomic mass of the ligand, correlate strongly with the binding affinity.

Our method is validated on two data-sets: compstatin analogs bound to C3, and fluorescein bound to antibody molecules. The two force-based descriptors, together with atomic masses of the ligands are apparently sufficient to capture both the enthalpic and entropic contributions to the free energy and therefore have great potential for binding affinity prediction. One limitation, however, is that these force-based descriptors require accurate ligand-macromolecule complex structures and will therefore need an x-ray crystallographic structure of at least one ligand-macromolecule complex.