The amino acid dice don't roll where they may

Weinreich and coworkers report in Science an interesting piece of molecular biology under the title "Darwinian Evolution Can Follow Only Very Few Mutational Paths to Fitter Proteins". They observed that, although there are 120 possible mutation paths for beta-lactamase to make the five amino acid changes that confer increased resistance on the host bacterium (i.e. any mutation first, any of the remaining four next, etc., so n=5!), 102 of those trajectories are closed because the interim steps offer no advantage or even are disadvantageous. They blame this on what they call "biophysical pleiotropy", meaning that the knock-on changes of one resistance-conferring mutation was often a less stable or imperfectly folded enzyme and this was not counterbalanced by most of the next changes in most processions of mutation. Their pithy sum-up refers to nicely to Stephen Jay Gould[52]:

[Our work] implies that the protein tape of life may be largely reproducible and even predictable.