Despite decades of research, the ability to generate effective antibodies completely on the computer is elusive, and antibody discovery and optimization continue to rely on iterative and time-consuming high-throughput screening. I will review the prerequisites for effective and universally applicable antibody design methods, the opportunities in developing methods for computational antibody optimization, and the underlying factors that explain why completely computational antibody design remains challenging.

Computational protein design has become a key tool in protein engineering. It is however clear that often computationally designed proteins are often suboptimal and require further optimization. Frequently, such optimization is performed by in vitro evolution techniques which are equally powerful and have been transformative in protein engineering. During my talk, I will discuss the strengths and weaknesses of each approach and their synergistic usage to solve hard problems in protein engineering that become accessible by the use of hybrid approaches. Lastly, I will discuss some of the potential applications of ML-driven approaches to enable both computational design and in vitro evolution.

The isolation of binding proteins from combinatorial libraries has typically relied on the use of a soluble, recombinantly expressed form of the target protein when performing magnetic selections or fluorescence-activated cell sorting. Appropriate target protein expression and subsequent purification represents a significant bottleneck in this process. As an alternative, the use of target proteins expressed on the surface of magnetized yeast cells in combinatorial library screening is discussed.

We have developed an approach that we call “systems biologics”, which combines large-scale systems biology with the development of new antibody drugs. The efficient pipeline extends from basic research through translational science, and it constitutes a new model for research and drug development. Through this model, cutting-edge systems biology basic research can be seamlessly translated into systems biologics: novel, multi-functional drugs, and diagnostics that take advantage of the complexities of human biology revealed by genomics data.

The growing need for antibodies with customized specificity provides a rich environment for engineering efforts. By leveraging the unique properties of neural networks, we developed a generative model for immunoglobulin 3D structures, with which diverse structures can be modeled with unprecedented speed. This "Generative Design" strategy explores dynamic structures and our preliminary experimental results on multiple targets support the plausibility of in silico design of epitope-specific antibodies.

Rapid allergic desensitization is a priority for emerging allergy therapies. Using yeast display and multi-parameter selections, we have isolated anti-IgEs that target and rapidly accelerate the dissociation of IgE from its high-affinity receptor. We then employ these new fast-acting anti-IgEs alongside engineered selection reagents to study the structure of the metastable pre-dissociation complex. Together these studies enable new therapeutic avenues for the anti-IgE inhibitor class.

Using the amber suppression-based noncanonical amino acid mutagenesis technique, we showed that a number of noncanonical amino acids with unique chemical functionalities can be genetically encoded in phage-displayed peptides. These unique chemical functionalities allowed simultaneous cyclization with a nearby cysteine for the generation of phage-displayed cyclic peptide libraries and directly targeting the active sites of therapeutic targets for improved and accelerated drug discovery. Successful demonstrations have been done with therapeutic targets including SIRT1, HDAC8, ENL, and BRD9. 

Chemoselective and site-selective modification of phage coat proteins allows facile incorporation of designer functional groups into phage-displayed peptide libraries, thereby extending the power of phage display to cover non-natural peptide libraries. This presentation will discuss our work on the development and evaluation of phage libraries that carry a reversible covalent warhead, which can greatly enhance a peptide's binding to its target protein. 

• Design and properties of lysine-targeting warheads
• Phage-displayed libraries of novel macrocyclic peptides
• Covalent "binding" peptide macrocycle as inhibitors for difficult-to-target proteins

bioPROTACs are expressable protein-based degraders that redirect ubiquitination to a target of interest. The design rules for such agents have yet to be fully worked out – what binder properties (affinity, epitope, stability) are critical, what linker properties (rigidity, length) are best, and which E3 ligase adaptors drive the greatest degree of target degradation? And in lieu of hard rules, what is the most efficient strategy for bespoke bioPROTAC optimization?

An unbiased discovery of proximity-dependent degraders and stabilizers will be reviewed and characterization of unexpected protein degraders and stabilizers will be discussed. The functional assessment of 300 E3 ligases and 50 DUBs for proximity-dependent degradation and stabilization will be covered and discovery of highly potent and broadly active E3 ligases for targeted protein degradation will be reviewed.

Bicycle Therapeutics is developing a unique class of chemically synthesized medicines based on its proprietary bicycle peptide (Bicycle) phage display platform. These bicyclic peptides, or Bicycles, are a class of highly constrained peptides characterized by the formation of two loops when the linear peptide is cyclized around a trivalent scaffold. Bicycles have broad utility and may confer several advantages over existing modalities – namely their small size, modularity for conjugation, and favorable pharmacokinetics. Bicycles are currently being explored in the clinic as Bicycle toxin conjugates (BTCs) for targeted delivery of cytotoxic payloads into tumors, and Bicycle tumor-targeted immune cell agonists (Bicycle TICAs). In the talk, I will share our research progress in the development of Bicycle conjugates.

The DARPin technology enables the exploration of new therapeutic designs on multiple disease pathways. Based on DARPin properties (small size, high affinity, excellent stability), we can generate single-domain DARPin agents for fast-in/fast-out radio ligand therapies (RLT) with high tumor accumulation – OR – generate multi-specific DARPin therapeutics such as MP0533, a half-life extended CD3-based T cell engager targeting simultaneously CD33, CD123, and CD70 to selectively target malignant AML cancer cells.

Single-domain antibody fragments known as VHH have emerged in the pharmaceutical industry as useful biotherapeutics. Here we describe the development of synthetic VHH libraries and a phage-to-yeast affinity maturation workflow for rapid in vitro selection of high affinity single domain binders. The phage VHH libraries have 10^11 unique CDRH3 diversity. After 2 rounds of phage panning, the yeast display libraries are then built to introduce additional CDRH1 and CDRH2 diversity. High affinity VHH binders are selected from yeast libraries. We show this platform is able to generate high affinity and potent antibodies with broad sequence diversity and epitope coverage.