Phi value analysis is an experimental protein engineering method used to study the structure of the folding transition state in small protein domains that fold in a two-state manner. Since the folding transition state is by definition a transient and partially unstructured state, its structure is difficult to determine by traditional methods such as protein NMR or X-ray crystallography. In phi-value analysis, the folding kinetics and conformational folding stability of the wild-type protein are compared with those of one or more point mutants. This comparison yields a phi value that seeks to measure the mutated residue's energetic contribution to the folding transition state (and thus the degree of native structure around the mutated residue in the transition state) from the relative free energies of the unfolded state, the folded state and the transition state for the wild-type and mutant proteins.

Typically, a high fraction of the protein's residues are mutated one by one to identify clusters of residues that are well-ordered in the folded transition state. The interactions of these residues can be validated using double-mutant-cycle phi analysis, in which the effects of the single mutants are compared with those of the double mutant. In general, the mutations are conservative and replace the original residue with a smaller one (cavity-creating mutations), most commonly alanine; however, others such as tyrosine-to-phenylalanine, isoleucine-to-valine and threonine-to-serine mutations are also used. Examples of proteins that have been studied by phi value analysis include chymotrypsin inhibitor, SH3 domains, individual domains of proteins L and G, ubiquitin, and barnase.

Phi value analysis fundamentally assumes a close relationship between structure and energy. If the energy landscape has a well-defined and relatively deep global minimum, the resemblance of a folding intermediate structure to the native state may closely correlate with the energy of that structure. However, if the energy landscape is relatively flat or has many local minima, the relationship may not hold strongly enough for free energy destabilizations to provide useful structural information. The method also assumes that the folding pathway is not significantly altered, although the folding energies may be. For nonconservative mutations this assumption might be fundamentally flawed; thus conservative substitutions are preferred, though they may yield smaller energetic destabilizations that are thus more difficult to detect experimentally. Lastly, the restriction of the phi value range as necessarily nonnegative assumes that the introduction of a mutation will not increase the stability and thus lower the energy of either the native or the transition state relative to those of the wild-type protein. Also, it is implicitly assumed that the interactions that stabilize a folding transition state are native-like in nature. Many recent studies of protein folding, however, have suggested that stabilizing non-native interactions in a folding transition state may aid in folding. An elegant example of this is given in Zarrine-Afsar et al. (2008) PNAS, where authors have demonstrated that stabilizing non-native interaction in the Fyn SH3 domain actually accelerated the folding rate of this protein.