Although the molecular mechanism and thermodynamic profile of a wide variety of chemical agents have been examined intensively in the past decades in terms of specific recognition of their protein receptors, to date the physicochemical nature of DNA-drug recognition and association still remains largely unexplored. The present study focused on understanding the structural basis, energetic landscape, and biological implications underlying the binding of small-molecule ligands to their cognate or non-cognate DNA receptors. First, a new method to capture the structural features of DNA-drug complex architecture was proposed and then used to correlate the extracted features with binding affinity of the complexes. By employing this method, a statistical regression-based predictor was developed to quantitatively evaluate the interaction potency of drug compounds with DNA in a fast and reliable manner. Subsequently, we use the predictor to examine the binding behavior of a number of structure-available, affinity-known DNA-drug complexes as well as a large pool of randomly generated DNA decoys in complex with the same drugs. It was found that (1) as compared with protein-DNA recognition, small-molecule agents have relatively low specificity in selecting their cognate DNA targets from the background of numerous random decoys; (2) the abundance of A-T base pairs in the DNA core motif exhibits a significant positive correlation with the affinity of drug ligand binding to the DNA receptor; and (3) high affinity seems not to be closely related to high selectivity for a DNA-targeting drug, although high-affinity drug entities have a greater possibility of being ranked computationally as top binders. We hope that this work will provide a preliminary insight into the molecular origin of sequence-specific interactions in DNA-drug recognition.