The study of noncovalent interactions, notably including drug-protein binding, relies heavily on the language of localized functional group contacts:h ydrogen bonding, p-p interactions, CH-p contacts, halogen bonding, etc. Applyingt he state-of-the-art functional group symmetry-adapted perturbation theory (F-SAPT) to an important question of chloro versus methyl aryl substitution in factor Xa inhibitor drugs, we find that al ocalized contact model provides an incorrect picture for the origin of the enhancement of chloro-containing ligands. Instead, the enhancement is found to originate from many intermediate-range contacts distributed throughout the binding pocket, particularly including the peptideb onds in the protein backbone. The contributionsf rom these contacts are primarily electrostatic in nature,b ut requirea bi nitio computations involvingn early the full drug-protein pocket system to be accurately quantified.Computational drug design,d espite severaln otable achievements, [1] remains af rontier area for theoretical chemistry,w ith numerous unresolved challenges. Formally,t he disciplines of electronic structure theory and statistical mechanics can provide arbitrarily accurate predictions of desired observables, e.g.,o ft he ligand-protein binding energy DG bind .H owever,i n practice, the computational effort required to accurately encapsulate the physics for these observablesi si ntractable. [2] For example, to accurately compute DG bind for ag iven ligand-protein system,one must implicitly or explicitly capture the contributions from the gas-phaseinteraction energy DE int ,solventeffects, including the desolvation penalties of the ligand and binding pocket, and the dynamic vibrational and conformational corrections. [2,3] Moreover, all of thesee ffects shouldb em odeled using an ab initio qualityp otential surface. Though much progress toward accurate characterization of DG bind has been made along the lines of quantum chemistry [4] and free energy perturbation theory, [3,5,6] direct quantification of this metric remains elusive.An alternative approach involves startingf rom ak nown and well-characterized ligand-protein system, and then proposing substitutions to the ligand to enhance the binding affinity.T his approachi so stensibly simpler, as one only needs to predict the binding energyd ifference DDG bind accurately.F or simple substitutions involving similarp olarities, the differences in ligand desolvation penalties and dynamical contributionsm ay be small relative to DDG bind .I ns uch cases, DDG bind will be largely governed by the differenceo ft he interaction energy DDE int between the ligand and the protein, and it suffices to characterize the latter quantity to predict, at least qualitatively, the substituted ligand's affinity.N ote that there exist many cases in which the differentiald esolvation penalties and dynamicalc ontributions are not small, and require direct computation of DDG bind for an accurate result (e.g.,i nt he case of boundl igandsw hich are partially exposed ...