Antibody Pharmacokinetics (PK) Prediction

The single most critical parameter governing IgG half-life. The neonatal Fc receptor (FcRn) binds antibodies in the acidic endosome (pH < 6.5) and releases them at physiological pH (7.4). ML models predict the KD ratio at pH 6.0 vs pH 7.4 to estimate recycling efficiency.

• Key residues: Ile253, His310, His435 in the CH2-CH3 domain interface
• Mutations like M252Y/S254T/T256E (YTE) increase affinity ~10-fold at pH 6.0
• Predicted using physics-based free energy perturbation (FEP) or graph neural networks trained on SPR data

Antibody clearance through renal filtration and lymphatic convection is inversely correlated with size. ML models use Stokes radius (Rh) and molecular weight as primary inputs for predicting systemic clearance rates.

• Full IgG (~150 kDa, Rh ~5.5 nm) resists renal filtration
• Fragments (Fab ~50 kDa, scFv ~25 kDa) clear rapidly via glomerular filtration
• Fc fusion proteins and PEGylation increase hydrodynamic radius to extend half-life
• Predicted from sequence using 3D structure prediction or empirical scaling laws

The isoelectric point (pI) and surface charge patch distribution influence non-specific tissue uptake and pinocytosis rate. ML models trained on PK data show that antibodies with pI > 8.5 exhibit faster systemic clearance due to enhanced electrostatic interactions with negatively charged cell surfaces.

• Fc engineering to lower pI (e.g., N434S mutation) reduces clearance
• Positive charge patches in CDRs accelerate hepatic sinusoidal endothelial cell uptake
• Predicted via Poisson-Boltzmann electrostatic calculations or sequence-based pI estimators

A saturable clearance pathway where antibody elimination is driven by binding to its pharmacological target. ML models must account for target expression levels, internalization rate, and binding affinity to predict non-linear PK profiles.

• High-affinity binding to rapidly internalizing membrane targets accelerates clearance
• TMDD dominates at low doses; FcRn-mediated recycling dominates at high doses
• Physiologically-based pharmacokinetic (PBPK) models integrate TMDD with tissue-level target expression data
• Predicted using antibody-antigen docking and internalization rate classifiers

Fc glycosylation at Asn297 modulates both effector function and PK. High-mannose glycans accelerate clearance via mannose receptor-mediated uptake in the liver, while afucosylated glycans enhance ADCC but may alter FcRn binding kinetics.

• Terminal sialylation extends half-life by masking galactose residues recognized by the asialoglycoprotein receptor (ASGPR)
• ML models predict glycan occupancy and composition from sequence context and cell line metadata
• Glycoengineering (e.g., CHO cell line engineering) is a key half-life optimization lever

Antibodies with low conformational stability (Tm < 60°C) are prone to unfolding and irreversible aggregation, which triggers rapid immune complex clearance and anti-drug antibody (ADA) responses. ML models predict Tm, Tagg, and aggregation hotspots from sequence and structure.

• Spatial aggregation propensity (SAP) scores identify hydrophobic patches driving self-association
• CDR loop flexibility correlates with aggregation risk
• Developability indices combine stability, solubility, and polyspecificity predictions into a single PK-relevant score
• Predicted using molecular dynamics simulations or deep learning stability classifiers

The primary mechanism governing the long half-life of IgG antibodies. The neonatal Fc receptor (FcRn) binds to the antibody's Fc region in the acidic environment of the endosome (pH < 6.5), rescuing it from lysosomal degradation and recycling it back to the cell surface, where it is released at physiological pH (7.4).

• pH-Dependent Binding: The interaction must be strong at acidic pH and weak at neutral pH for efficient recycling.
• Key Residues: Histidine residues at the CH2-CH3 domain interface (e.g., H310, H435) are critical for this pH switch.
• Competition: High concentrations of endogenous IgG can saturate FcRn, accelerating the clearance of a therapeutic antibody.

A nonlinear clearance pathway where the antibody's pharmacological target itself drives elimination. When the antibody binds to its membrane-bound or soluble target, the complex is internalized and degraded. This pathway is saturable.

• Impact on PK: Causes dose-dependent half-life; clearance is faster at low doses when target is abundant and slows at higher doses when target is saturated.
• Modeling: TMDD models incorporate binding affinity (Kd), target expression levels, and internalization rates.
• Prediction: AI models must account for target turnover rates to accurately forecast PK profiles.

An immune response where the patient develops antibodies against the therapeutic antibody, forming immune complexes that are rapidly cleared. This is a major cause of unpredictable PK and loss of efficacy.

• Clearing ADAs: Bind to the drug and accelerate its elimination, drastically reducing half-life.
• Neutralizing ADAs: Block the drug's binding to its target, neutralizing its pharmacological activity.
• Prediction: Computational immunogenicity prediction tools identify T-cell epitopes in the antibody sequence that may trigger ADA formation.

A mechanistic modeling approach that mathematically represents the body as a series of interconnected compartments (organs and tissues) connected by blood flow. It simulates drug concentration-time profiles using physiological parameters.

• Parameters: Incorporates tissue volumes, blood flow rates, FcRn expression levels, and antibody-specific properties like hydrodynamic radius and charge.
• Species Translation: Enables scaling of PK predictions from preclinical species (mouse, monkey) to humans.
• AI Integration: Machine learning can parameterize PBPK models by predicting tissue distribution coefficients from antibody sequence and structure.

The biophysical properties of an antibody that govern its extravasation from blood vessels into tissues and its renal filtration. Larger hydrodynamic radius and negative charge generally reduce clearance.

• Renal Filtration Cutoff: Molecules with a hydrodynamic radius above ~5 nm (the size of an IgG) are largely excluded from glomerular filtration.
• Electrostatic Repulsion: The negatively charged glycocalyx of the glomerular basement membrane repels anionic molecules, reducing filtration.
• Engineering: Modifying the variable domain's surface charge through computational design can modulate tissue penetration and half-life.

An advanced antibody engineering strategy that extends half-life by making antigen binding conditional on pH. The antibody binds tightly to its target at neutral pH but releases it in the acidic endosome.

• Mechanism: After internalization, the antibody releases the antigen in the endosome, allowing the free antibody to bind FcRn and be recycled, while the antigen is degraded.
• Histidine Scanning: Introducing histidine residues into the CDRs creates a pH-sensitive switch due to histidine's pKa (~6.0).
• AI Prediction: Models predict mutations that introduce pH-dependent binding without sacrificing affinity.