Machine Learning Feature Selection for Predicting High Concentration Therapeutic Antibody Aggregation.

Protein aggregation can hinder the development, safety and efficacy of therapeutic antibody-based drugs. Developing a predictive model that evaluates aggregation behaviors during early stage development is therefore desirable. Machine learning is a widely used tool to train models that predict data with different attributes. However, most machine learning techniques require more data than is typically available in antibody development. In this work, we describe a rational feature selection framework to develop accurate models with a small number of features. We applied this framework to predict aggregation behaviors of 21 approved monospecific monoclonal antibodies at high concentration (150 mg/mL), yielding a correlation coefficient of 0.71 on validation tests with only two features using a linear model. The nearest neighbors and support vector regression models further improved the performance, which have correlation coefficients of 0.86 and 0.80, respectively. This framework can be extended to train other models that predict different physical properties.

A single molecular descriptor to predict solution behavior of therapeutic antibodies.

Despite the therapeutic success of monoclonal antibodies (mAbs), early identification of developable mAb drug candidates with optimal manufacturability, stability, and delivery attributes remains elusive. Poor solution behavior, which manifests as high solution viscosity or opalescence, profoundly affects the developability of mAb drugs. Using a diverse dataset of 59 mAbs, including 43 approved products, and an array of molecular descriptors spanning colloidal, conformational, charge-based, hydrodynamic, and hydrophobic properties, we show that poor solution behavior is prevalent (>30%) in mAbs and is singularly predicted (>90%) by the diffusion interaction parameter (k D), a dilute-solution measure of colloidal self-interaction. No other descriptor, individually or in combination, was found to be as effective as k D. We also show that well-behaved mAbs, a substantial subset of which bear high positive charge and pI, present no disadvantages with respect to pharmacokinetics in humans. Here, we provide a systematic framework with quantitative thresholds for selecting well-behaved therapeutic mAbs during drug discovery.

Frustration and folding of a TIM barrel protein.

Triosephosphate isomerase (TIM) barrel proteins have not only a conserved architecture that supports a myriad of enzymatic functions, but also a conserved folding mechanism that involves on- and off-pathway intermediates. Although experiments have proven to be invaluable in defining the folding free-energy surface, they provide only a limited understanding of the structures of the partially folded states that appear during folding. Coarse-grained simulations employing native centric models are capable of sampling the entire energy landscape of TIM barrels and offer the possibility of a molecular-level understanding of the readout from sequence to structure. We have combined sequence-sensitive native centric simulations with small-angle X-ray scattering and time-resolved Förster resonance energy transfer to monitor the formation of structure in an intermediate in the Sulfolobus solfataricus indole-3-glycerol phosphate synthase TIM barrel that appears within 50 µs and must at least partially unfold to achieve productive folding. Simulations reveal the presence of a major and 2 minor folding channels not detected in experiments. Frustration in folding, i.e., backtracking in native contacts, is observed in the major channel at the initial stage of folding, as well as late in folding in a minor channel before the appearance of the native conformation. Similarities in global and pairwise dimensions of the early intermediate, the formation of structure in the central region that spreads progressively toward each terminus, and a similar rate-limiting step in the closing of the ß-barrel underscore the value of combining simulation and experiment to unravel complex folding mechanisms at the molecular level.

A Hyperthermophilic Phage Decoration Protein Suggests Common Evolutionary Origin with Herpesvirus Triplex Proteins and an Anti-CRISPR Protein.

Virus capsids are protein shells that protect the viral genome from environmental assaults, while maintaining the high internal pressure of the tightly packaged genome. To elucidate how capsids maintain stability under harsh conditions, we investigated the capsid components of the hyperthermophilic phage P74-26. We determined the structure of capsid protein gp87 and show that it has the same fold as decoration proteins in many other phages, despite lacking significant sequence homology. We also find that gp87 is significantly more stable than mesophilic homologs. Our analysis of the gp87 structure reveals that the core "ß tulip" domain is conserved in trimeric capsid components across numerous double-stranded DNA viruses, including Herpesviruses. Moreover, this ß barrel domain is found in anti-CRISPR protein AcrIIC1, suggesting a mechanism for the evolution of this Cas9 inhibitor. Our work illustrates the principles for increased stability of gp87, and extends the evolutionary reach of the ß tulip domain.

Interplay between drying and stability of a TIM barrel protein: a combined simulation-experimental study.

Recent molecular dynamics simulations have suggested important roles for nanoscale dewetting in the stability, function, and folding dynamics of proteins. Using a synergistic simulation-experimental approach on the αTS TIM barrel protein, we validated this hypothesis by revealing the occurrence of drying inside hydrophobic amino acid clusters and its manifestation in experimental measures of protein stability and structure. Cavities created within three clusters of branched aliphatic amino acids [isoleucine, leucine, and valine (ILV) clusters] were found to experience strong water density fluctuations or intermittent dewetting transitions in simulations. Individually substituting 10 residues in the large ILV cluster at the N-terminus with less hydrophobic alanines showed a weakening or diminishing effect on dewetting that depended on the site of the mutation. Our simulations also demonstrated that replacement of buried leucines with isosteric, polar asparagines enhanced the wetting of the N- and C-terminal clusters. The experimental results on the stability, secondary structure, and compactness of the native and intermediate states for the asparagine variants are consistent with the preferential drying of the large N-terminal cluster in the intermediate. By contrast, the region encompassing the small C-terminal cluster experiences only partial drying in the intermediate, and its structure and stability are unaffected by the asparagine substitution. Surprisingly, the structural distortions required to accommodate the replacement of leucine by asparagine in the N-terminal cluster revealed the existence of alternative stable folds in the native basin. This combined simulation-experimental study demonstrates the critical role of drying within hydrophobic ILV clusters in the folding and stability of the αTS TIM barrel.