We aim to improve the design of synthetic libraries for molecular recognition scaffolds using fibronectin type III as a model system. Specifically, two factors were investigated: tailored diversity at expected paratope positions and wild-type bias at structurally critical positions. It was hypothesized that the amino acid distribution in antibody complementarity-determining regions is a better basis for paratope diversity than either an approximated equal distribution (such as NNB) or a narrow distribution (such as serine/tyrosine). Moreover, we demonstrate how this distribution can be achieved inexpensively through designed degenerate codons. Secondly, we implement wild-type bias at structurally critical or non-exposed positions to reduce the entropic cost of binding and focus diversity at positions with more potential impact on binding. Such positions are identified through analysis of stability, solvent accessibility, and phylogenetic sequences.

These design elements were incorporated into a fibronectin type III domain library in the yeast surface display platform. Stability analysis validated inclusion of wild-type bias at multiple positions yet demonstrated flexibility at other sites. Selection and maturation of binders towards multiple targets yielded clones with picomolar to nanomolar affinities. Direct competition of this library versus NNB and serine/tyrosine libraries without structural bias yielded preferential selection of binders from the new library thereby validating its improved design. Sequence analysis as well as binding and stability assays of binders and their mutants are providing insight for further refinement of library design.

Extended Abstract Status: Not Uploaded