Behavioural Profiling of Therapeutic Antibodies via Non-equilibrium Interfacial Wave Dynamics

Abstract

Therapeutic antibody performance depends not only on sequence and structure, but also on how the molecule responds to physico-chemical perturbations encountered in its environment. Whereas sequence and static structure can be inferred under controlled conditions, behaviour is a conditional, multidimensional response to environment, most directly characterised by applying defined perturbations and quantifying the resulting dynamics. Conventional early-stage developability methods capture isolated dimensions of this response under near-equilibrium conditions, deferring integrated behavioural assessment to late-stage characterisation, when the cost of correction is highest. We introduce Variations in Interfacial Behaviour under Excitation (VIBE), an interfacial wave method implemented via Liquid State Intelligence (LSI), a sensing architecture that transduces molecular perturbations at the air-liquid interface into high-dimensional wave patterns. A colloidal liquid substrate operated near a thermodynamic transition couples small molecular perturbations to large dynamical responses, integrating structural flexibility, charge distribution, and surface hydrophobicity into a single behavioural readout from microgram-scale samples. Applied to antibodies previously characterised by industrial benchmarks, the primary behavioural descriptor, VIBE1, functioned as a high-precision triage tool, flagging candidates carrying multiple biophysical liabilities. In a clinical-stage cohort, the proportion of high-VIBE1 antibodies declined progressively from early trials through approval, and high-VIBE1 candidates showed an elevated clinical failure rate. Concordance analysis against estab-lished methods confirmed that VIBE1 captures a composite signal spanning hydrophobicity, polyreactivity, self-interaction, and thermal stability rather than recapitulating any single conventional readout. These findings establish interfacial wave sensing as a low-material modality for early-stage developability assessment, repositioning molecular behaviour from late-stage validation to discovery-phase characterisation.