In this work we report on generating/using simulated metabolite basis sets for the quantification of in vivo MRS signals, assuming that they have been acquired by using the PRESS pulse sequence. To that end we have employed the classes and functions of the GAMMA C++ library. By using several versions of our PRESS-simulation program, we were able to study the single-voxel selection, required for detecting in vivo MRS signals. Furthermore, by introducing in one of the versions a modified spatial summation scheme, that comes down to crusher-gradient averaging, we could realize a decrease in computation time by about a factor of 256. We have used four different simulated metabolite basis sets in the quantification of a real-world 3T human-brain 1H MRS signal. The best quantification is obtained, when including into the simulation program -as closely as possible- the related details of the PRESS-based single-voxel selection