Introduction

Peptide drug-likeness refers to the set of physicochemical and structural properties that predict whether a peptide can be developed into a viable therapeutic agent. Unlike small molecules, peptides occupy a unique chemical space between traditional drugs and biologics. This article examines the rules governing peptide drug-likeness, predictors of oral bioavailability, and strategies to enhance membrane permeation.

Lipinski’s Rule of Five for Peptides

Original Rules

Lipinski’s Rule of Five (Ro5) predicts oral bioavailability for small molecules:

• Molecular weight ≤ 500 Da
• LogP ≤ 5
• Hydrogen bond donors ≤ 5
• Hydrogen bond acceptors ≤ 10

Peptide Limitations

Most peptides violate Ro5:

• MW > 500 Da: Even dipeptides approach 200–300 Da
• HBD > 5: Each amide bond contributes 1 NH
• HBA > 10: Each amide bond contributes 1 C=O
• PSA > 140 Ų: Peptides have extensive polar surface

Extended Rule of Five (eRo5)

For peptides, the extended rules are more relevant:

• MW ≤ 1000 Da (or ≤ 1500 Da with modifications)
• cLogP ≤ 5
• PSA ≤ 200 Ų (≤ 140 Ų for good permeability)
• Number of rotatable bonds ≤ 15
• Number of amide bonds ≤ 8

Oral Bioavailability Predictors

Physicochemical Parameters

Parameter	Optimal Range	Impact on Absorption
MW	500–1500 Da	Permeability decreases above 1000 Da
LogP	1–5	Optimal for passive diffusion
PSA	70–140 Ų	>140 Ų limits permeability
HBD	≤ 5	Reduces permeability
HBA	≤ 10	Reduces permeability
Rotatable bonds	≤ 10	Flexibility reduces permeability

Veber Rules

For oral bioavailability, Veber’s rules apply:

• Rotatable bonds ≤ 10
• PSA ≤ 140 Ų

PSA Calculations

Polar surface area (PSA) is a critical predictor:

• PSA < 70 Ų: Good oral bioavailability
• PSA 70–140 Ų: Moderate oral bioavailability
• PSA > 140 Ų: Poor oral bioavailability

PSA contributions by functional group:

Group	PSA (Ų)
Amide (secondary)	26
Amide (primary)	43
Carboxylic acid	37
Hydroxyl	20
Amine (primary)	26
Guanidinium	43

Permeation Mechanisms

Passive Transcellular Diffusion

The primary route for peptide absorption:

1. Partitioning from aqueous phase into lipid bilayer
2. Diffusion through membrane interior
3. Partitioning back into aqueous phase

Requirements:

• Adequate lipophilicity (LogP 1–5)
• Low PSA (< 140 Ų)
• Conformational flexibility (membrane适应ation)

Paracellular Transport

Transport through tight junctions between enterocytes:

• Limited to small peptides (< 200–300 Da)
• Highly dependent on concentration gradient
• Regulated by tight junction proteins

Carrier-Mediated Transport

Active transport via peptide transporters:

• PepT1 (SLC15A1): Broad substrate specificity
• PepT2 (SLC15A2): Higher affinity, lower capacity
• PAT1 (SLC36A1): Small amino acids, D-amino acids

PepT1 substrates:

• Di- and tripeptides
• β-Lactam antibiotics
• ACE inhibitors
• Protease inhibitors

Transcytosis

Receptor-mediated endocytosis:

• Transferrin receptor
• Folate receptor
• LDL receptor

Strategies for Enhanced Oral Bioavailability

Molecular Modifications

1. N-methylation:

• Reduces HBD count
• Enhances membrane permeability
• Examples: Cyclosporin A (7 N-methylations)

2. D-amino acid substitution:

• Alters conformation
• Resists proteolysis
• Enhances permeability

3. Cyclization:

• Reduces conformational entropy
• Masks HBD/HBA
• Enhances permeability

4. Lipid conjugation:

• Fatty acid acylation
• Cholesterol conjugation
• Enhances membrane interaction

Formulation Strategies

1. Permeation enhancers:

• Salicylates: Tight junction modulation
• Bile salts: Membrane solubilization
• Surfactants: Transcellular transport

2. Enzyme inhibitors:

• Protease inhibitors: Reduce degradation
• Example: Aprotinin, soybean trypsin inhibitor

3. Nanoparticle delivery:

• Liposomes: Membrane fusion
• Polymeric nanoparticles: Endocytosis
• Solid lipid nanoparticles: Lymphatic uptake

Computational Predictors

Machine Learning Models

1. Random Forest classifiers:

• Training data: Known oral peptides
• Features: MW, LogP, PSA, HBD, HBA
• Accuracy: 70–85%

2. Support Vector Machines:

• Non-linear classification
• Kernel functions for complex feature spaces
• Accuracy: 75–90%

3. Deep neural networks:

• Convolutional neural networks for sequence features
• Recurrent neural networks for sequence patterns
• Accuracy: 80–95%

Molecular Dynamics Predictions

1. Membrane permeability simulations:

• Peptide partitioning into lipid bilayer
• Free energy profiles
• Diffusion coefficients

2. Conformational analysis:

• Membrane-active conformations
• Hydrogen bonding patterns
• Amphipathicity in membrane environment

Case Studies

Cyclosporin A

Properties:

• MW: 1202 Da
• LogP: 2.92
• PSA: 280 Ų
• Oral bioavailability: ~30%

Design features:

• 7 N-methylations (reduce HBD)
• 7 D-amino acids (resist proteolysis)
• Cyclic structure (constrain conformation)
• Lipophilic side chains (enhance permeability)

Linaclotide

Properties:

• MW: 1323 Da
• Oral bioavailability: ~5–10%
• Mechanism: Gut-restricted

Design features:

• 3 disulfide bonds (constrain structure)
• Minimal systemic absorption
• Local action in GI tract

Semaglutide (Oral)

Properties:

• MW: 4114 Da
• Oral bioavailability: ~1%
• Enhancement: SNAC co-formulation

Design features:

• N-terminal modification (resist DPP-IV)
• Albumin binding (extend half-life)
• SNAC absorption enhancer

Lipinski Rule Violations and Solutions

Common Violations

Violation	Peptide Example	Solution
MW > 500	Most peptides	Minimize length
HBD > 5	Linear peptides	N-methylation, cyclization
HBA > 10	Linear peptides	Cyclization
PSA > 140	Linear peptides	N-methylation, lipidation

Successful Violations

Some peptides succeed despite Ro5 violations:

• Cyclosporin A: MW 1202 Da
• Vancomycin: MW 1449 Da
• Ramoplanin: MW 1880 Da

Bioavailability Enhancement Technologies

Chemical Enhancement

1. Prodrugs:

• Ester prodrugs: Increase lipophilicity
• Phosphate prodrugs: Increase solubility
• Amino acid prodrugs: Exploit PepT1

2. Peptide mimetics:

• β-peptides: Resist proteolysis
• Peptoids: N-substituted glycines
• Peptide isosteres: Non-cleavable mimics

Physical Enhancement

1. Permeation enhancers:

• EDTA: Calcium chelation
• Sodium caprate: Tight junction modulation
• Cell-penetrating peptides: Transcytosis

2. Formulation technologies:

• enteric coating: Protect from gastric acid
• Mucoadhesive systems: Increase residence time
• Nanoparticles: Enhance uptake

Predictive Models

Quantitative Structure-Permeability Relationships (QSPR)

Model:

log(Perm) = a·MW + b·LogP + c·PSA + d·HBD + e·HBA + intercept

Typical coefficients:

• MW: negative (larger = less permeable)
• LogP: positive (up to optimal)
• PSA: negative (more polar = less permeable)
• HBD: negative (more H-bonds = less permeable)

Machine Learning Predictions

Features for ML models:

1. Physicochemical: MW, LogP, PSA, charge
2. Sequence-based: Composition, motifs
3. Structural: Secondary structure, amphipathicity
4. Descriptors: E-state, topological indices

Summary

Peptide drug-likeness requires balancing multiple physicochemical properties to achieve adequate oral bioavailability. While traditional Lipinski rules provide guidance, peptides require modified criteria and specialized strategies. The most successful approaches combine molecular modifications (N-methylation, cyclization, D-amino acids) with formulation technologies (permeation enhancers, nanoparticles). Computational prediction models continue to improve, enabling more rational design of orally bioavailable peptides.

Deep dive: Explore Peptide Modification Strategies for detailed modification techniques, or read about Peptide ADMET for comprehensive pharmacokinetic analysis.

Test yourself: Take the Peptide Drug-Likeness Quiz or study with Drug-Likeness Flashcards.