In this paper, we discussed our recent experience with the use of computational modeling tools in studying the binding interaction of small molecular weight ligands with their protein targets. Specific examples discussed here include the interaction of estrogens with human protein disulfide isomerase (PDI) and its pancreas-specific homolog (PDIp), and the interaction of dietary flavonoids with human cyclooxygenase (COX) I and II. Using human PDIp as an example, biochemical analysis revealed that the estrogen-binding activity is only associated with PDIp's b-b´ domain combination but not associated with the single b or b´ domain or any other domains. Homology modeling was then used to build a threedimensional structure of the human PDIp's b-b´ fragment. Docking analyses predicted that a hydrogen bond, formed between the 3-hydroxyl group of estradiol and His278 of PDIp's E2-binding site, is critical for the binding interaction. This binding model was then experimentally confirmed by a series of experiments, such as selective mutations of the predicted binding site amino acid residues and the selective modifications of the functional groups of the ligands. Similar combinatorial approaches were used successfully to identify the binding site structure of human PDI for estradiol and the binding site structures of human COX I and II for their phenolic co-substrates. The success with these combinatorial approaches provides the basis for using computational modeling-guided approaches in characterizing the ligand binding site structures of complex proteins whose structures are difficult to decipher with crystallographic studies.