Quantitative structure-property relationship (QSPR) models were developed to correlate physicochemical properties of structurally unrelated drugs with extent of in vitro binding to colesevelam, and predicted values were compared with drug exposure changes in vivo following coadministration. The binding of 17 drugs to colesevelam was determined by an in vitro dissolution drug-binding assay. Data from several clinical studies in healthy volunteers to support administration of colesevelam in diabetic patients were also collected along with existing in vivo literature data and compared with in vitro results. Steric, electronic, and hydrophobic descriptors were calculated for test compounds, and univariate and partial least squares regression approaches were used to derive QSPR models to evaluate which of the molecular descriptors correlated best with in vitro binding. A quadrant analysis evaluated the correlation between predicted/actual in vitro binding results and the in vivo data. The in vitro binding assay exhibited high sensitivity, identifying those compounds with a low probability of producing relevant in vivo drug interactions. Drug lipophilicity was identified as the primary determinant of in vitro binding to colesevelam by the final univariate and partial least squares models (R(2) = 0.69 and 0.98; Q(2) = 0.48 and 0.59). The in vitro assay and in silico models represent predictive tools that may allow investigators to conduct only informative clinical drug interaction studies with colesevelam.