Triglyceride composition of rat liver was analyzed by HPLC with a non-linear elution gradient of acetone in acetonitrile using a light-scattering detector. Triglyceride molecular species were predicted by means of simple and multiple linear regression analysis between the log k' of the HPLC peaks and molecular variables, equivalent carbon number (ECN) of the possible triglycerides, and chain length (CL) and number of double bounds (DB) of each of the fatty acids in the glycerol molecule.