High levels of accuracy in transcription, aminoacylation of tRNA, and mRNA translation are essential for all life forms. However, high accuracy also necessarily means large energy dissipation and slow kinetics. Therefore, in vivo there is a fine tuned balance between rate and accuracy of key chemical reactions. We have shown that in our optimized in vitro bacterial protein synthesis system we have in vivo compatible rate and accuracy of ribosomal protein elongation. Our measurements of the temperature and the pH dependence of peptide bond formation with native substrates also suggest that the chemical step of peptidyl transfer, rather than tRNA accommodation, limits the rate of peptide bond formation. This work has made it possible to study ribosomal peptidyl transfer with native substrates. Furthermore, we have developed a general theoretical model for the rate-accuracy trade-off in enzymatic reactions. When considering this trade-off for protein synthesis in the context of the living bacterial cell, where cognate aa-tRNAs compete for ribosome binding with an excess of non-cognate aa-tRNAs, the model predicts an accuracy optimum where the inhibitory effect of non-cognate substrate binding and the efficiency loss due to high discard rate of cognate aa-tRNAs are minimized. However, these results also show that commonly used biochemical systems for protein synthesis studies operate at exceptionally suboptimal conditions. This makes it difficult, if not impossible, to relate the biochemical data to protein synthesis in the living cell. To validate our theoretical model we developed a method, based on variation of the concentration of Mg2+ ions in the buffer, to study the rate-accuracy trade-off of bacterial protein synthesis in vitro. We found a linear trade-off between rate and accuracy of tRNA selection on the ribosome, from which we could estimate the maximal accuracy. Exploiting this method for a complete set of single-mismatch readings by one tRNA species, we found simple patterns of genetic code reading, where the accuracy was highest for the second and lowest for the third codon position. The results bridge the gap between in vivo and in vitro protein synthesis and allow calibration of our test tube conditions to those of the living cell.