• Opening and preparing a complex • Annotating proteins • Calculating protein properties • Protein patches and the Patch Analyzer application • Analyzing protein contacts • Looking at the antigen surface – Molecular surface and electrostatic maps • Virtual mutagenesis with the Protein Builder and Protein Design applications

• Template and loop searching with the Antibody Modeler • Building a homology model of the Fv domain • Identifying and removing glycosylation sites • The Antibody Database Project Search panel – Viewing statistics about the antibody structures • Building a humanized model of the Fab domain

Product properties such as solubility, solution viscosity, and self-association states are all influenced by electrostatic and non-electrostatic protein-protein interactions that can potentially be engineered during candidate selection, or optimized during formulation development. Assessment of protein-protein interactions can be done both experimentally and computationally via a number of different techniques while trying to balance restrictions of material needs and time constraints. This talk will focus on established and emerging approaches to quantify protein-protein interactions for therapeutic proteins, from both experimental and computational perspectives.

Accurate modeling of antibody (Ab) structures is an integral aspect of biologics drug discovery and directly impacts key areas such as humanisation, affinity maturation and epitope mapping. Abs are unlike other proteins and as such the development of modeling tools specific for Abs will rely on a large and varied specific dataset of structural information to be available in the public domain. It is estimated that a typical human is capable of producing 10¹⁰ distinct Ab molecules. Since there are approximately 2600 Ab structures in the PDB to date and only about 1000 of these are unique from a sequence perspective, there are still considerable gaps in our structural knowledge of a typical human Ab repertoire which need to be filled. The benefit of additional structural knowledge can have an impact at three different scales: Atomic – where better modeling of a single given Ab in atomic detail will help understand and rationalise modifications to its molecular properties (binding / flexibility / stability / developability), Molecular – where better coverage of Ab structural space will increase the probability of being able to model (at all) any given sequence of interest and, Repertoire/Patient – where the ability to structurally classify entire Ab repertoires from, e.g. a patient blood sample, will enrich the information obtained from sequence alone and enable better and more detailed understanding of normal vs. disease states, biomarker development, patient to patient variability, etc. The main goal of our project is to increase the number of antibody structures deposited in the PDB and show that these have a positive impact on the ability to model unknown structures.

Early stage developability assessments of monoclonal antibody (mAb) candidates can help reduce risks and costs associated with their product development. Forecasting viscosity of highly concentrated mAb solutions is an important aspect of such developability assessments. Reliable predictions of concentration-dependent viscosity behaviors for mAb solutions in platform formulations can help screen or optimize drug candidates for flexible manufacturing and drug delivery options. Here, we present a computational method to predict concentration-dependent viscosity curves for mAbs solely from their sequence-structural attributes. This method was developed using experimental data on 16 different mAbs whose concentration-dependent viscosity curves were experimentally obtained under standardized conditions. Each concentration-dependent viscosity curve was fitted with a straight line, via logarithmic manipulations, and the values for intercept and slope were obtained. Intercept, which relates to antibody diffusivity, was found to be nearly constant. In contrast, slope, the rate of increase in solution viscosity with solute concentration, varied significantly across different mAbs, demonstrating the importance of intermolecular interactions toward viscosity. Next, several molecular descriptors for electrostatic and hydrophobic properties of the 16 mAbs derived using their full-length homology models were examined for potential correlations with the slope. An equation consisting of hydrophobic surface area of full-length antibody and charges on VH, VL, and hinge regions was found to be capable of predicting the concentration-dependent viscosity curves of the antibody solutions. Availability of this computational tool may facilitate material-free high-throughput screening of antibody candidates during early stages of drug discovery and development.

Computational modeling of protein-protein interactions is increasingly playing a more important role in the design and optimization of biologics. Computational methods for predicting protein-protein binding sites or epitopes have profound applications in a number of areas in the development of biologics, from understanding the mode of action to the modulation of protein properties. This work presents a novel algorithm for predicting likely antigen epitopes from protein-protein docking results using the MOE software platform. The approach generates an ensemble of poses which represent the most favorable interactions. Protein-protein residue contacts are then used to generate interaction fingerprints which serve to identify Boltzmann-weighted clusters of poses and extract consensus epitope residues. This method produces at least one predicted epitope with significant overlap to the native structure ranked in the top five clusters.

In the last few years, it has been shown that with the availabilities of large data and capabilities of higher sampling, modeling accuracy has improved in a number of applications. In homology modeling, using multiple input templates has improved modeling accuracies in both aligned and unaligned regions. In protein-protein interface designs, large sampling not only help improving binding affinities, but also specificities. And with increasing knowledge in epitopes, we can now develop a general approach for reducing immunogenicity.

Self-association behavior of monoclonal antibodies (MAb) are thought to modulate the high viscosity observed in their high concentration solutions. Computationally predicting the self-association behavior of concentrated antibodies based on their amino acid sequence is of significant interest but remains a challenge. Here we construct Y-shaped 10-site rigid Coarse Grained (CG) models of antibodies that interact via screened Coulombic electrostatic and 12-6 Lennard-Jones potentials. The point charges on each CG site are placed so that the key low order electrostatic multipole moments of the of full-length antibody charge distribution are optimally reproduced. The new CG model is applied to simulations of two therapeutic MAbs with experimentally different viscosity behavior, over a wide range of concentrations (60−200 mg/mL). We show that the calculated electrostatic multipole moments (up to the octupole order) for the two MAbs are significantly different, even though the only difference between the MAbs lies with a few amino acids at the antigen-binding loops (CDRs), which in turn leads to distinct charge placement for the corresponding CG models. Microsecond simulations using the new CG models reveal the formation of dense regions or clusters that vary in size depending on MAb type and solution concentration, consistent with experiment. The self-diffusion and total potential energies calculated at different concentrations are qualitatively consistent with the viscosity behavior of the two antibodies. Our CG model is particularly applicable as a screening tool for selecting MAbs with desirable viscosity properties early during lead candidate selection.

We describe technologies and strategies for streamlining creation of pharmaceutical products by integrating development (and manufacturing) together with discovery. In order to bring about this approach, new technologies are needed in addition to a new mindset. We describe both with a focus on technologies for mAbs, including formulation technologies. Target areas are aggregation and viscosity prediction and design, but we also discuss briefly, oxidation and hydrolysis. In addition, we make the case for investment in strategic approaches to developability.