Owing to a rapid rise in global energy demand in various sectors including power, agriculture and transport, there is a tremendous increase in demand for conventional diesel fuel. Biodiesel are emerging as renewable alternative to diesel with better emission characteristics (except nitric oxides). The biodiesel could be produced from various feedstock including vegetable oils, animal fats, algae, etc. and thus, vary significantly in their fatty acid methyl ester composition and physico-chemical properties and thereby, engine performance and emissions. In the present work, the effects of biodiesel compositional variations in conjunction with changes in engine load and injection timings are captured using a multi-linear regression model which is applied to predict performance and emission characteristics of a single cylinder diesel engine. The biodiesel compositional effects on engine performance and emissions are captured through two composition based parameters, viz. straight chain saturation factor (SCSF) and modified degree of unsaturation (DUm) which can be estimated directly from the measured fatty acid methyl ester composition of biodiesel. The developed model is capable of predicting engine performance and exhaust emission characteristics at a constant engine speed under varying injection timings and load conditions from no load to full load. For each of the operating conditions, a correlation matrix analysis is carried out to examine the significance of changes in biodiesel composition and it is observed that the composition effects are more pronounced near full load conditions. The predictions from the developed model are validated with measurements made in a single-cylinder, naturally aspirated, diesel engine fuelled with different biodiesel fuels of varying compositions. Furthermore, optimization studies on composition based parameters viz. SCSF and DUm are also carried out using genetic algorithm for the minimization of NOx and brake specific fuel consumption. The predictions are found to be in good agreement with a regression coefficient of above 0.9 and an absolute average deviation of less than 5% for all the investigated performance and emission parameters except smoke. Although the developed regression model is applicable for a particular engine type, similar approach can be extended to any engine type by suitably modifying the correlation coefficients. Thus, the proposed composition based regression model can predict engine characteristics with any biodiesel fuel type whose composition is known apriori. The usefulness of the proposed model is that it reduces the number of experiments required to be performed in any given engine with different biodiesel fuel types and also brings out the effects of biodiesel compositional variations on engine characteristics. The approach used for the formulation in the present work can further be utilized and tuned for any engine type.