Gas-phase optimization of single biological molecules and of small active-site biological models has become a standard approach in first principles computational enzymology. The important role played by the surrounding environment (solvent, enzyme, both) is normally only accounted for through higher-level single point energy calculations performed using a polarizable continuum model (PCM) and an appropriate dielectric constant with the gas-phase-optimized geometries. In this study we analyze this widely used approximation, by comparing gas-phase-optimized geometries with geometries optimized with different PCM approaches (and considering different dielectric constants) for a representative data set of 20 very important biological molecules--the 20 natural amino acids. A total of 323 chemical bonds and 469 angles present in standard amino acid residues were evaluated. The results show that the use of gas-phase-optimized geometries can in fact be quite a reasonable alternative to the use of the more computationally intensive continuum optimizations, providing a good description of bond lengths and angles for typical biological molecules, even for charged amino acids, such as Asp, Glu, Lys, and Arg. This approximation is particularly successful if the protonation state of the biological molecule could be reasonably described in vacuum, a requirement that was already necessary in first principles computational enzymology.