A decoupled alignment kernel for peptide membrane permeability predictions
8 affiliations
- 1Center for Applied Intelligent Systems Research in Health, Halmstad University, Kristian IV:s väg 3, 30118, Halmstad, Sweden. [email protected]
- 2Molecular AI, Discovery Sciences, R&D, AstraZeneca, Pepparedsleden 1, 43183, Mölndal, Sweden. [email protected]
- 3Division of Systems and Synthetic Biology, Department of Life Sciences, Chalmers University of Technology, Kemigården 1, 41296, Gothenburg, Sweden
- 4Molecular AI, Discovery Sciences, R&D, AstraZeneca, Pepparedsleden 1, 43183, Mölndal, Sweden
- 5Medicinal Chemistry, Research and Early Development, Respiratory & Immunology, BioPharmaceuticals R&D, AstraZeneca, Pepparedsleden 1, 43183, Mölndal, Sweden
- 6Center for Applied Intelligent Systems Research in Health, Halmstad University, Kristian IV:s väg 3, 30118, Halmstad, Sweden
- 7Department of Research and Development (FoU), Region Halland, Södra vägen 9, 30118, Halmstad, Sweden
- 8Centre for Environmental and Climate Science, Computational Science for Health and Environment, Lund University, Kontaktvägen 10, 22362, Lund, Sweden
Abstract
Cyclic peptides are promising modalities for targeting intracellular sites; however, cell-membrane permeability remains a key bottleneck, exacerbated by limited public data and the need for well-calibrated uncertainty. Instead of relying on data-eager complex deep learning architecture, we propose a monomer-aware decoupled global alignment kernel (MD-GAK), which couples chemically meaningful residue-residue similarity with sequence alignment while decoupling local matches from gap penalties. MD-GAK is a relatively simple kernel. To further demonstrate the robustness of our framework, we also introduce a variant, PMD-GAK, which incorporates a triangular positional prior. As we will show in the experimental section, PMD-GAK can offer additional advantages over MD-GAK, particularly in reducing calibration errors. Since our focus is on uncertainty estimation, we use Gaussian Processes as the predictive model, as both MD-GAK and PMD-GAK can be directly applied within this framework. We demonstrate the effectiveness of our methods through an extensive set of experiments, comparing our fully reproducible approach against state-of-the-art models, and show that it outperforms them across all metrics.Scientific contributionWe introduce monomer-aware decoupled global alignment kernels for Gaussian processes (MD-GAK and position-aware PMD-GAK) that align cyclic peptides at the sequence level using chemically rich monomer fingerprints and explicit positional priors, yielding positive-definite similarity measures tailored to permeability modeling. Compared with order-agnostic fingerprint methods, standard global-alignment kernels and state-of-the-art graph and language-model baselines, our alignment-aware GPs provide improved discrimination, probabilistic calibration and scaffold-level robustness under stringent, leakage-controlled cyclic-peptide permeability benchmarks.
Related
- CIPER: Interpretable Permeability Predictions Guide Cyclic Peptide Editing
- Predicting the Fourth Analogue: Learning Stereochemical Transfer in Cyclic-Peptide Permeability
- The Same Edit, a Different Outcome: Relational Learning for Cyclic Peptide Permeability
- Fast, lightweight BBB permeability predictions for molecular design
- Charge Ratio and Distribution Effects on Lys/Ala Peptide Conformation and Adsorption onto the Lipid Bilayer Membrane (POPS)
BibTeX
@article{amirahmadi2026decoupled,
title = {{A decoupled alignment kernel for peptide membrane permeability predictions}},
author = {Ali Amirahmadi and Gökçe Geylan and Leonardo De Maria and Farzaneh Etminani and Mattias Ohlsson and Alessandro Tibo},
journal = {Journal of Cheminformatics},
year = {2026},
month = aug,
doi = {10.1186/s13321-026-01276-5},
url = {https://doi.org/10.1186/s13321-026-01276-5},
note = {CC BY},
}
RIS
TY - JOUR
TI - A decoupled alignment kernel for peptide membrane permeability predictions
AU - Ali Amirahmadi
AU - Gökçe Geylan
AU - Leonardo De Maria
AU - Farzaneh Etminani
AU - Mattias Ohlsson
AU - Alessandro Tibo
PY - 2026
DA - 2026/08/04
JO - Journal of Cheminformatics
DO - 10.1186/s13321-026-01276-5
UR - https://doi.org/10.1186/s13321-026-01276-5
AB - Cyclic peptides are promising modalities for targeting intracellular sites; however, cell-membrane permeability remains a key bottleneck, exacerbated by limited public data and the need for well-calibrated uncertainty. Instead of relying on data-eager complex deep learning architecture, we propose a monomer-aware decoupled global alignment kernel (MD-GAK), which couples chemically meaningful residue-residue similarity with sequence alignment while decoupling local matches from gap penalties. MD-GAK is a relatively simple kernel. To further demonstrate the robustness of our framework, we also introduce a variant, PMD-GAK, which incorporates a triangular positional prior. As we will show in the experimental section, PMD-GAK can offer additional advantages over MD-GAK, particularly in reducing calibration errors. Since our focus is on uncertainty estimation, we use Gaussian Processes as the predictive model, as both MD-GAK and PMD-GAK can be directly applied within this framework. We demonstrate the effectiveness of our methods through an extensive set of experiments, comparing our fully reproducible approach against state-of-the-art models, and show that it outperforms them across all metrics.Scientific contributionWe introduce monomer-aware decoupled global alignment kernels for Gaussian processes (MD-GAK and position-aware PMD-GAK) that align cyclic peptides at the sequence level using chemically rich monomer fingerprints and explicit positional priors, yielding positive-definite similarity measures tailored to permeability modeling. Compared with order-agnostic fingerprint methods, standard global-alignment kernels and state-of-the-art graph and language-model baselines, our alignment-aware GPs provide improved discrimination, probabilistic calibration and scaffold-level robustness under stringent, leakage-controlled cyclic-peptide permeability benchmarks.
SN - 1758-2946
ER -