Developability-by-design for bispecific antibodies in discovery: combining computational prediction and early screening for candidate selection.

Abstract

INTRODUCTION The molecular and architectural complexity of bispecific antibodies can introduce developability risks beyond those encountered with conventional monospecific antibodies. Therefore, whether developability should be integrated into the discovery process rather than used as a late-stage filter is an increasingly important question. AREAS COVERED This structured narrative review examines literature identified in PubMed/MEDLINE, Embase, Scopus and Web of Science from database inception, supplemented by citation tracking. Molecular engineering, sequence-, structure-, and machine-learning-based prediction, architecture-related, inherited, and emergent liabilities, and material-efficient tests for assembly, stability, and pharmacokinetic risk are all assessed. For candidate ranking and nomination, a four-tier design-predict-screen-redesign methodology that is iterative and rooted in the goal product profile is suggested. EXPERT OPINION Developability across bispecific formats cannot be captured by a single computational score or individual assay. Rather, computational predictions should serve to generate testable hypotheses, while candidate nomination requires verified molecular identity, correctly assembled yield, orthogonal testing of purified constructs, and validation under conditions directly relevant to the formulation, administration route, and mechanism of action. Ultimately, field-wide advancement hinges on format-diverse datasets, assay-resolved labels, identity-aware external validation, and prospective evidence.