Evaluation of Cell Permeability in Macrocyclic Peptides: An Exploratory Computational Chemistry Approach
(2025) In Master’s Theses in Mathematical Sciences BERM06 20251Mathematics (Faculty of Sciences)
- Abstract
- Macrocyclic peptides (MCPs) are a promising drug modality for targeting protein–protein interactions, offering high selectivity, metabolic stability, and potential cell permeability. However, predicting their membrane permeability remains challenging due to their conformational flexibility, environment-dependent polarity, and ability to adopt chamaleonic behavior, e.g. by changing the number of intramolecular hydrogen bonds. These features, combined with their frequent violation of Lipinski’s Rule of Five, make traditional small-molecule models poorly suited for MCPs. This exploratory thesis investigates whether computational tools can accurately predict permeability-relevant properties of MCPs and how well these predictions align with... (More)
- Macrocyclic peptides (MCPs) are a promising drug modality for targeting protein–protein interactions, offering high selectivity, metabolic stability, and potential cell permeability. However, predicting their membrane permeability remains challenging due to their conformational flexibility, environment-dependent polarity, and ability to adopt chamaleonic behavior, e.g. by changing the number of intramolecular hydrogen bonds. These features, combined with their frequent violation of Lipinski’s Rule of Five, make traditional small-molecule models poorly suited for MCPs. This exploratory thesis investigates whether computational tools can accurately predict permeability-relevant properties of MCPs and how well these predictions align with experimental data. Machine learning models were developed using the CycPeptMPDB database and trained on both 2D and 3D molecular descriptors. A curated subset of 60 matched MCP pairs, differing slightly in structure but showing significant differences in permeability, was analyzed to evaluate descriptor–permeability relationships. To bridge the gap between computational predictions and experimental validation, two matched pairs (four peptides) were synthesized. Their logP values were determined using the shake-flask method, and their conformations were analyzed in polar and apolar solvents using NMR spectroscopy. These experimental data, provided by colleagues at the company, were compared with structures generated from molecular dynamics (MD) simulations. The results showed that key descriptors such as logP, PSA, and the presence of intramolecular hydrogen bonds strongly correlate with permeability, but also that static descriptor calculations often fail to capture stereochemical effects or dynamic folding behavior. MD simulations and structural shape metrics provided more nuanced insights into conformational differences affecting permeability. Overall, this exploratory study demonstrates that integrating machine learning, dynamic modeling, and experimental validation offers a powerful approach for understanding and predicting the permeability of MCPs, advancing their rational design in the beyond-rule-of-five chemical space. (Less)
- Popular Abstract (Swedish)
- Makrocykliska peptider är en ny typ av läkemedelskandidater som kan påverka mål i kroppen som vanliga läkemedel ofta inte kommer åt – till exempel när två proteiner ska interagera. Trots att dessa peptider är relativt stora och komplexa, har vissa visat sig kunna tas upp i kroppen vid tablettintag, något som annars är ovanligt för större molekyler. Men det är fortfarande oklart vilka egenskaper som gör att vissa av dessa peptider kan ta sig igenom cellens skyddande membran och bli aktiva i kroppen. Syftet med det här explorativa examensarbetet var att undersöka om datorbaserade metoder kan hjälpa till att förutsäga vilka makrocykliska peptider som har god förmåga att passera cellmembran. För att göra detta analyserades nästan 7000 olika... (More)
- Makrocykliska peptider är en ny typ av läkemedelskandidater som kan påverka mål i kroppen som vanliga läkemedel ofta inte kommer åt – till exempel när två proteiner ska interagera. Trots att dessa peptider är relativt stora och komplexa, har vissa visat sig kunna tas upp i kroppen vid tablettintag, något som annars är ovanligt för större molekyler. Men det är fortfarande oklart vilka egenskaper som gör att vissa av dessa peptider kan ta sig igenom cellens skyddande membran och bli aktiva i kroppen. Syftet med det här explorativa examensarbetet var att undersöka om datorbaserade metoder kan hjälpa till att förutsäga vilka makrocykliska peptider som har god förmåga att passera cellmembran. För att göra detta analyserades nästan 7000 olika peptider med hjälp av maskininlärning och molekyldynamiska simuleringar. Datorn tränades att hitta samband mellan olika egenskaper, som fettlöslighet, förmåga att binda till vatten (polär yta) och inre vätebindningar, och hur lätt peptiderna tar sig in i celler. För en mer detaljerad analys valdes 60 peptidpar ut där varje par bestod av två nästan identiska molekyler med olika förmåga att ta sig in i celler. Det visade sig att små strukturella förändringar kunde påverka hur peptiderna veckade sig och hur de interagerade med sin omgivning, faktorer som i sin tur påverkar deras upptag i kroppen. Fyra av peptiderna tillverkades i labb av kollegor på företaget och testades med experimentella metoder, där resultaten stämde väl överens med datorns förutsägelser. Slutsatsen är att en kombination av datorberäkningar och experimentella tester kan ge en bättre förståelse för hur framtidens läkemedel kan utformas för att fungera effektivt i kroppen, även när de bryter mot de klassiska reglerna för läkemedelsdesign. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9200398
- author
- Rasul, Honia LU
- supervisor
- organization
- course
- BERM06 20251
- year
- 2025
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- Descriptors, Machine learning, Macrocyclic peptides, Membrane permeability, Molecular dynamics
- publication/series
- Master’s Theses in Mathematical Sciences
- report number
- LUNFTB-3001-2025
- ISSN
- 1404-6342
- other publication id
- 2025:E55
- language
- English
- id
- 9200398
- date added to LUP
- 2025-09-02 16:55:01
- date last changed
- 2025-09-02 16:55:01
@misc{9200398,
abstract = {{Macrocyclic peptides (MCPs) are a promising drug modality for targeting protein–protein interactions, offering high selectivity, metabolic stability, and potential cell permeability. However, predicting their membrane permeability remains challenging due to their conformational flexibility, environment-dependent polarity, and ability to adopt chamaleonic behavior, e.g. by changing the number of intramolecular hydrogen bonds. These features, combined with their frequent violation of Lipinski’s Rule of Five, make traditional small-molecule models poorly suited for MCPs. This exploratory thesis investigates whether computational tools can accurately predict permeability-relevant properties of MCPs and how well these predictions align with experimental data. Machine learning models were developed using the CycPeptMPDB database and trained on both 2D and 3D molecular descriptors. A curated subset of 60 matched MCP pairs, differing slightly in structure but showing significant differences in permeability, was analyzed to evaluate descriptor–permeability relationships. To bridge the gap between computational predictions and experimental validation, two matched pairs (four peptides) were synthesized. Their logP values were determined using the shake-flask method, and their conformations were analyzed in polar and apolar solvents using NMR spectroscopy. These experimental data, provided by colleagues at the company, were compared with structures generated from molecular dynamics (MD) simulations. The results showed that key descriptors such as logP, PSA, and the presence of intramolecular hydrogen bonds strongly correlate with permeability, but also that static descriptor calculations often fail to capture stereochemical effects or dynamic folding behavior. MD simulations and structural shape metrics provided more nuanced insights into conformational differences affecting permeability. Overall, this exploratory study demonstrates that integrating machine learning, dynamic modeling, and experimental validation offers a powerful approach for understanding and predicting the permeability of MCPs, advancing their rational design in the beyond-rule-of-five chemical space.}},
author = {{Rasul, Honia}},
issn = {{1404-6342}},
language = {{eng}},
note = {{Student Paper}},
series = {{Master’s Theses in Mathematical Sciences}},
title = {{Evaluation of Cell Permeability in Macrocyclic Peptides: An Exploratory Computational Chemistry Approach}},
year = {{2025}},
}