Antibody-Drug Conjugate (ADC) Design

The computational identification of optimal amino acid residues on the antibody surface for payload attachment. The goal is to identify sites that are solvent-accessible for efficient conjugation chemistry but do not interfere with antigen binding or Fc receptor interactions.

• Engineered Cysteine Residues: Algorithms predict sites where introduced cysteines will remain reduced for maleimide conjugation without disrupting disulfide bond integrity
• Solvent Accessible Surface Area (SASA): Molecular dynamics simulations calculate the exposure of candidate residues to ensure chemical reactivity
• Paratope Proximity Exclusion: Models automatically exclude residues within a defined radius of the complementarity-determining regions (CDRs) to preserve affinity

Computational modeling of the distribution and homogeneity of the Drug-to-Antibody Ratio, a critical quality attribute that directly impacts pharmacokinetics, toxicity, and efficacy. The goal is to achieve a uniform DAR, typically 2, 4, or 8, rather than a heterogeneous mixture.

• Stochastic Conjugation Modeling: Monte Carlo simulations predict the statistical distribution of DAR species based on reaction kinetics and residue reactivity
• Hydrophobicity Thresholding: Machine learning models predict the hydrophobicity increase caused by linker-payload attachment to avoid aggregation at higher DARs
• Bystander Killing Radius: Computational pharmacodynamics models correlate DAR with the predicted diffusion distance of released payload for solid tumor penetration

Quantum mechanical and molecular dynamics simulations that predict the chemical stability of the covalent linker connecting the antibody to the cytotoxic payload in both circulatory and intracellular environments.

• Plasma Stability Half-Life: Models predict the rate of premature payload release in systemic circulation, a primary cause of off-target toxicity
• Cathepsin B Cleavage Kinetics: Docking simulations predict the susceptibility of protease-cleavable linkers to lysosomal enzymes for efficient intracellular payload release
• pH-Responsive Linker Design: pKa prediction algorithms model the protonation state of acid-labile linkers to ensure stability at physiological pH 7.4 and rapid hydrolysis in endosomal pH 5.0

Computational prediction of a released payload's ability to diffuse across cell membranes to kill neighboring antigen-negative tumor cells, a phenomenon critical for treating tumors with heterogeneous target expression.

• LogP and Membrane Permeability: Quantitative structure-property relationship (QSPR) models predict the octanol-water partition coefficient to assess cell-permeability of the free payload
• Charge State at Lysosomal pH: pKa calculators determine whether the released payload will be neutral and membrane-permeable or charged and trapped within the target cell
• Tumor Penetration Depth: Spatial agent-based models simulate payload diffusion gradients within a 3D tumor spheroid to predict the effective killing radius

The use of antibody homology models to predict the structural context of conjugation sites when a high-resolution crystal structure is unavailable. This is essential for early-stage ADC candidate screening.

• Framework Region Conservation Analysis: Algorithms leverage the high structural conservation of IgG framework regions to transfer conjugation site annotations from known structures
• Local Microenvironment Assessment: Models predict the electrostatic and steric environment around a candidate conjugation site to anticipate its impact on linker-payload stability
• Aggregation Propensity After Conjugation: Spatial aggregation propensity (SAP) algorithms recalculate the antibody surface hydrophobicity map after in silico attachment of the hydrophobic linker-payload

Integrated computational models that simulate the full systemic journey of an ADC, from intravenous infusion to intratumoral payload release, to predict the therapeutic index.

• Deconvoluted Clearance Pathways: Models separately track the clearance rates of the intact ADC, the unconjugated antibody, and the free payload
• Tumor Uptake and Penetration: Physiologically-based pharmacokinetic (PBPK) models incorporate tumor blood flow, vascular permeability, and antigen density to predict intratumoral ADC accumulation
• Minimum Effective Concentration (MEC): Simulations predict the threshold intratumoral payload concentration required for tumor stasis, directly informing clinical dose selection

The computational identification of optimal amino acid residues on the antibody for linker attachment. The goal is to identify sites that do not interfere with antigen binding or structural integrity.

• Engineered Cysteine Residues: Algorithms predict solvent-accessible positions for introduced cysteines to achieve a uniform Drug-to-Antibody Ratio (DAR).
• Solvent Accessibility: Models calculate the Solvent Accessible Surface Area (SASA) of candidate residues to ensure linker accessibility.
• Structural Stability: Predicts whether a conjugation site will destabilize the antibody's tertiary structure or CH2 domain.

In silico modeling of the chemical linker that connects the cytotoxic payload to the antibody. The linker must be stable in systemic circulation but efficiently cleaved inside the target cell.

• Cleavable Linkers: Models simulate the pH-dependent hydrolysis of hydrazone linkers in acidic endosomes (pH 5.0-6.0) or the proteolytic cleavage of valine-citrulline dipeptides by cathepsin B.
• Non-Cleavable Linkers: Predicts the stability of thioether bonds that require complete lysosomal degradation of the antibody for payload release.
• Bystander Effect: Predicts the hydrophobicity of the released payload to determine its ability to cross cell membranes and kill neighboring antigen-negative tumor cells.

The computational selection and optimization of the cytotoxic warhead. Modern ADCs use payloads with picomolar potency that target fundamental cellular processes.

• DNA-Damaging Agents: Models predict the minor groove binding affinity of calicheamicin or the cross-linking potential of pyrrolobenzodiazepine (PBD) dimers.
• Microtubule Inhibitors: Predicts the binding pose of maytansinoid derivatives (DM1, DM4) to the vinca binding site on tubulin.
• Topoisomerase I Inhibitors: Docking simulations for camptothecin analogs like exatecan and deruxtecan to the DNA-topoisomerase I cleavage complex.

The computational prediction of a released payload's ability to diffuse out of the target cell and into neighboring tumor cells. This property is critical for treating tumors with heterogeneous antigen expression.

• Hydrophobicity LogP: Models predict the octanol-water partition coefficient of the free payload. A moderate LogP facilitates membrane permeability.
• Charge State: Predicts the ionization state of the payload at physiological and endosomal pH, which dictates its ability to cross lipid bilayers.
• Diffusion Gradients: Spatial models simulate the concentration gradient of the released payload within a solid tumor microenvironment.

The computational generation of a full 3D model of the ADC complex to assess structural integrity and antigen accessibility. This integrates the antibody, linker, and payload into a single model.

• Antibody Framework: Uses template-based modeling from known IgG structures to build the constant and variable domains.
• Linker Conformational Sampling: Molecular dynamics simulations sample the vast conformational space of the flexible linker to predict its average solution state.
• Payload Masking: Predicts whether the attached payload is sterically shielded by the antibody surface, which can prevent premature interaction with off-target cell membranes.

The use of physiologically-based pharmacokinetic (PBPK) models to simulate the absorption, distribution, metabolism, and excretion of the ADC. A key challenge is modeling the dynamic equilibrium between the intact ADC and the free payload.

• Deconjugation Rate: Predicts the rate of linker-payload hydrolysis in plasma, a primary driver of systemic toxicity.
• FcRn-Mediated Recycling: Models the pH-dependent binding of the ADC's Fc region to the neonatal Fc receptor, which extends circulatory half-life.
• Tumor Penetration: Simulates the slow diffusion and binding-site barrier effect that limits ADC penetration into solid tumors.