Stealth Biologics

Biotherapeutics, including antibodies, enzymes, signaling peptides, and toxins, are revolutionizing disease therapy, but these next-generation drugs carry an inherent risk of eliciting detrimental immune responses. Stealth’s proprietary technology seamlessly integrates advanced computational models, protein design algorithms, protein engineering methodologies, and immunoassays to develop optimized therapeutic candidates that are less immunogenic but functionally equivalent to, or better than, nature’s own biomolecules.

Stealth Biologics® employs a cutting-edge protein design platform that leverages advanced computational methods to develop stable, active, and immunotolerant therapeutic candidates. The primary driver of a typical anti-biotherapeutic immune response is molecular recognition of a protein’s constituent T cell epitopes. We render a therapeutic protein more "stealthy" by mutating it so as to eliminate such immunogenic peptides, while ensuring that the therapeutic function is not compromised.

Three key aspects, elaborated below, distinguish the Stealth approach. We simultaneously optimize immunogenicity and function. We assess impacts of combinations of mutations on a protein as a whole. We identify globally optimal and near-optimal designs. Our platform thus enables us to engineer portfolios of highly active biotherapeutic candidates with minimized effector T cell recognition profiles.

In order to assess the effects of possible mutations on function and immunogenicity, we leverage computational models that integrate evolutionary information and/or structure-based energy evaluation with sequence-based T cell epitope prediction. We design either a set of functionally deimmunized variants, ranging over aggressiveness and mutational loads, or a deimmunized library, enriched in a diverse set of functionally deimmunized clones. Our novel combinatorial optimization algorithms generate designs that make the best trade offs between the assessments of immunogenicity and function. In a closed-loop approach, experimental results from initial assays are fed back into the design process in order to improve and focus further refinements.


In order for the target protein to maintain stability and activity, deimmunizing mutations must not adversely affect interactions among its amino acids. We therefore model and optimize the effects of mutations on the full protein, rather than treating individual constituent peptides as isolated entities.


The wild-type amino acid sequence is in some sense naturally optimized, and mutations are likely to be detrimental, moreso for epitope-deleting mutations that may require substitutions for hydrophobic residues. Thus we explicitly model and optimize trade offs between immunogenicity and maintenance of function, and select a diverse portfolio of engineered variants across the Pareto frontier of undominated designs.


The sequence space of possible designs presents a very rugged landscape over which to optimize, especially when considering dual objectives for immunogenicity and function. Thus we employ advanced combinatorial optimization methods that guarantee globally optimal and near-optimal solutions, thereby avoiding dead-end paths that limit more conventional, myopic strategies.