Developability
Assessment

A protein therapeutic can bind its target with picomolar affinity and still fail in development if it cannot be manufactured reliably. Developability refers to the collection of biophysical and biochemical properties that determine whether a molecule can be expressed at scale, purified efficiently, formulated at high concentration, and stored with acceptable shelf life. The key parameters include expression yield, solubility, aggregation propensity, viscosity at high concentration, thermal stability, chemical stability (deamidation, oxidation, isomerization), and immunogenicity risk.

Historically, developability issues surfaced late in development—often during manufacturing scale-up or formulation studies—when changing the molecule meant restarting years of work. The modern approach is to assess these properties during lead selection, so that candidates entering development already meet minimum thresholds for each critical attribute.

Expression and solubility are the first gatekeepers. A candidate that expresses below 0.5 g/L in CHO or aggregates above 5% after Protein A purification is unlikely to survive process development. Aggregation is particularly insidious—it reduces yield, triggers immunogenic responses, and complicates formulation. Techniques like size-exclusion chromatography (SEC), dynamic light scattering (DLS), and differential scanning fluorimetry (DSF) quantify these risks early.

For subcutaneous delivery, viscosity at high concentration (typically 100–200 mg/mL) becomes critical. Antibodies with excessive charge patches or hydrophobic surface regions often exhibit non-Newtonian viscosity that makes them impossible to inject through standard gauge needles. Immunogenicity, assessed through T-cell epitope prediction and comparison to human germline sequences, determines whether the therapeutic will elicit anti-drug antibodies (ADAs) that neutralize efficacy over time.

Computational tools can flag many developability liabilities from sequence and structure alone, before any protein is expressed. Surface hydrophobicity patches, charge asymmetry, unpaired cysteines, N-linked glycosylation motifs in CDRs, deamidation-prone asparagine-glycine motifs, and methionine oxidation sites can all be predicted in silico. These flags do not replace experimental characterization, but they allow you to deprioritize high-risk candidates before committing synthesis resources.

When applied at the design stage—scoring thousands of computational candidates for both binding and developability simultaneously—this approach shifts the quality of your lead panel dramatically. Instead of selecting the tightest binders and hoping they behave, you select from candidates that are already predicted to express well, remain monomeric, and avoid known sequence liabilities.