DNA gets mutated in every cell of every organism. Some mutations cause a substitution in the string of amino acids that fold into a protein. This could make no difference, or it coud be deadly.
Consider an enzyme that gets mutated in a bacterial cell. Having evolved over millions of years in growth competition with other microbes, this bacterium’s enzymes are finely tuned, and a random mutation is probably going to lower the enzyme’s efficiency, or prevent it from working at all.
This does not necessarily mean the bacterium will be outcompeted and die. In good times, organisms are generally more than capable of survival, even with some defects. This results in a “fitness saturation” effect, in which only particularly bad mutations have any impact on the organism. However, the reason less harmful mutations are mutations, and not the default sequence in that genome, is that in times of environmental stress, every detail matters.
This stress-dependent fitness saturation is beautifully illustrated by Stiffler, Hekstra, and Ranganathan, who modeled mutation impact on a bacterial enzyme as a function of the effect on the enzyme’s kinetics. This antibiotic resistance enzyme could handle a range of mutations when it didn’t have much to do, but higher concentrations of antibiotic put more stress on the enzyme, and otherwise neutral mutations became detrimental.
Many multicellular organisms have two or more versions of every chromosome per cell, and this takes the saturation effect even further. Now if a protein is nonfunctional due to an inherited mutation, there’s another copy in every cell, so the organism might be fine as long as the nonfunctional copy doesn’t get in the way. This is the basis of recessive alleles.
So far we’ve covered two levels of mutation impact, but for multicellular organisms there is a third level, and its relationship to the second level is even more complicated than that between the first and second. Many mutations that are damaging to a cell have no effect on the organism because the cell simply dies and is replaced (unless it’s a germline mutation, in which case all cells have the mutation). Mutations that are a minor nuisance to a cell might persist, and only after steady accumulation in different genes would the cumulative effect cause problems. Finally, mutations that increase cell growth may be a welcome optimization in unicellular organisms, but can be among the most dangerous of mutations in us multicellular folk if they manage to escape cellular safeguards (with help from some other mutations) and grow into a tumor.
Without getting into the methods we currently use to predict the fitness impact of protein variants, it doesn’t seem like there is any easy shortcut. Accurate predictions will require detailed knowledge of each protein’s function, atom by atom, in its full range of natural environments. In other words, I think we’ll need detailed physical simulations of proteins in their cellular environments before we can get anything more than vague predictions.