Antibodies are Y-shaped, versatile protein structures able to recognize with high affinity and specificity the full range of antigens (from small molecules to proteins) that an organism may encounter. This property has been leveraged in experimental binding assays such as ELISA and ELISPOT and in many life-saving medications (i.e. Avastin, Rituximab, Herceptin, etc.). Traditional antibody design methods rely on successive steps of library construction, mutagenesis and screening with limited computationally derived input. However, antibodies are an excellent target for computational design due to their function that is typically limited to binding, not catalysis, and the availability of well-established rules linking their primary sequence to structure. The team has already developed, published, and made freely available the OptCDR method for the de novo design of antibody binding pockets composed by the Complementarity Determining Regions (CDRs) against any specified antigen. Using OptCDR as a starting point, a workflow for the de novo design of the entire variable regions of fully-human antibodies to bind the desired epitope of any specified antigen will be pursued. The developed methods will be experimentally validated by designing antibody libraries to bind CD20, a peptide antigen that is a therapeutically relevant target in B-cell lymphomas and leukemias. The experimental component of this project will enable the fine-tuning of the proposed computational workflow and the quantitative assessment of the efficacy of the design predictions.
Under Aim 1, a combinatorial database of all germline antibody variable domain structures encoded in the human genome will be constructed. Using this combinatorial database of germline structures, the development and dissemination of the computational tools necessary for the de novo design of fully human antibodies against any specified antigen will be pursued under Aim 2. The envisioned computational design method OptMAVEn, (Optimal Method for Antibody Variable region Engineering) will expand the concepts pioneered in OptCDR to the design of the entire variable domains instead of only the CDRs. This will be followed by the design, construction and screening of five anti-CD20 antibody libraries to test the effectiveness of computations to drive antibody design (Aim 3). The five libraries, with the same approximate size of 5*108 antibodies, will progressively explore bolder computational-derived modifications. They will span random mutagenesis, saturation mutagenesis of six rationally chosen positions, targeted computational redesign of twelve positions, selection of entirely new CDRs with OptCDR, and de novo design of the entire variable domains with OptMAVEn.
Beyond the methodological advances, the comprehensive experimental characterization will provide a standard for fairly evaluating the performance of computations in antibody design, potentially revealing key benefits and inadequacies in modeling and simulation. The lessons learned in this study would be broadly applicable in other protein design endeavors. On the educational front, undergraduate students will be introduced to scientific research. All research results and methods will be broadly disseminated through journal publications, conference presentations, course-work and by making available through the web all developed software programs, databases, experimental results and protocols.
