Protein Engineering Protocols - Mycobacteriology research center

132 Denault and PelletierSubheading 3.1.2. regarding calculation of library representation). This allowsonly the partial exploration of a completely randomized 10-residue peptidelibrary (or a protein encoding 10 fully randomized residues), far from allowingfor the complete exploration of an average protein. This implies that there aretwo courses of action in creating protein libraries. In the first course, the targetprotein can be “fully” randomized; in this case, the library obtained representsa subset of all of the possibilities (see Subheading 1.2.). This can be extremelyinformative if knowledge regarding the relative importance of the wild-typesequence for a given function is required. Those residues that are statisticallyless-often mutated in the selected proteins are inferred to be critical for retentionof the characteristic that is selected for; this inference is analogous to thatused in determining functional residues by multiple alignments of naturalsequences. In the second course of action, the randomization can be limited;only specific residues (contiguous or not) are fully randomized and/or a bias isintroduced into the randomization to limit the number of different possibilities.This can result in a completely defined library of known size; the size can thenbe kept within the constraints described in Subheading 1.1. The library size canalso be matched to further constraints if screening of all library members, ratherthan sampling of a subset, is considered important.1.2. Library RepresentationInherent to the calculation of the required library size is the consideration ofthe library representation that one needs to achieve for each specific application.For a library selection to be efficient, the number of independent variantsencoded must be appropriate to the selection capacity of the strategy used. Asdescribed in Subheading 1.1., the creation of extremely large and diverse DNAlibraries is no longer a challenge. Indeed, one must generally restrict the size ofthe library to that which can adequately be screened. The tools for selection ofa desired phenotype (while maintaining a linkage to the genotype) have beenmore difficult to develop than those for library creation because they are not sobroadly applicable as the tools for manipulating DNA. Thus, a diversity ofselections strategies must be developed to select proteins having diverse functions(refer to ref. 2).In general, manual selection allows the experimenter to screen hundreds tothousands of protein variants; automation can increase the range by orders ofmagnitude. In vivo selection can allow for rapid screening of thousands tobillions of variants. Cell-free systems may be the most powerful experimentaltechnique (excluding computational approaches), allowing the researcher to screenup to 10 14 protein variants. Clearly, the number of variants treated by any ofthese means is extremely limited relative to the potential variety of protein variantsone could theoretically create, emphasizing the requirement for matching