Incorporating consideration for safety issues while selecting solvents for processes has become crucial in light of the chemical process accidents involving solvents that have taken place in recent years. Computer aided molecular design (CAMD) is a methodology that has been researched recently for designing compounds with required target properties and can be applied for selection of safer solvents as well. An important aspect of this methodology concerns the prediction of properties given the structure of the molecule. This paper utilizes one such emerging method for prediction of a hazardous property, flash point, which is indicative of the flammability of solvents. Quantitative structure property relationship (QSPR) and topological indices have been used in this paper to predict flash point properties of different classes of solvents. Multiple linear regression and back-propagation neural network analysis were used to model the flash point. The neural network model showed higher accuracy (training set, r = 0.948, R-2 = 0.898). However, there are certain limitations associated with using QSPR in CAMD which have been discussed and need further work. This paper advances the "forward problem" of CAMD using QSPR which has not been researched extensively in the past.