CycDelta: Multimodal Residue-Graph Learning of Permeability Changes in Cyclic Peptides
- 1Cyclon Therapeutics
Abstract
Membrane permeability remains a major challenge in cyclic-peptide drug discovery, where optimization often depends on estimating changes relative to measured reference compounds. Existing predictors mainly model absolute permeability, while heterogeneity across literature sources complicates pooled training and comparative evaluation. We present CycDelta, a multimodal residue-graph model trained directly on within-source permeability differences. CycDelta combines directed message passing with pretrained Uni-Mol monomer embeddings, physicochemical descriptors and assay information in a shared encoder for reference and query peptides. To evaluate this comparative formulation under controlled conditions, we establish CycDeltaBench, comprising in-distribution, scaffold-based out-of-distribution, external and zero-shot unseen-assay tests. On the out-of-distribution task, CycDelta achieved Pearson 0.69 versus 0.46 for the strongest baseline. It also led all evaluated methods on two external data sets and achieved zero-shot Spearman 0.70 versus 0.32 for the strongest baseline on an unseen assay. Residue-level attribution relates predictions to monomer contributions, while gradient-guided screening supports prioritization of permeability-enhancing analogue hypotheses. These results position CycDelta as a transferable and interpretable framework for comparative permeability prediction and cyclic-peptide optimization.
Related
- Prioritizing peptides for targeted mass spectrometry experiments using deep learning
- AntiOxiPred: Predicting Antioxidant Peptides using Machine Learning Techniques and BLAST
- A multimodal deep learning for rice crop health monitoring using vegetation index feature selection
- CUBE: Multimodal Representation Learning Reveals Biological Structure Across Histomorphology, Spatial Protein Phenotypes, and Transcriptome-Associated Signals
- Simultaneous multi-task assessment of five neurological language disorders via cross-corpus multimodal learning
BibTeX
@misc{qian2026cycdelta,
title = {{CycDelta: Multimodal Residue-Graph Learning of Permeability Changes in Cyclic Peptides}},
author = {J. Qian and Y. Zhou and Y. Cui and Z. Gao and L. Wu and S. Li and M. Yang and D. Wang},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.18.752569},
url = {https://www.biorxiv.org/content/10.64898/2026.09.18.752569v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - CycDelta: Multimodal Residue-Graph Learning of Permeability Changes in Cyclic Peptides
AU - J. Qian
AU - Y. Zhou
AU - Y. Cui
AU - Z. Gao
AU - L. Wu
AU - S. Li
AU - M. Yang
AU - D. Wang
PY - 2026
DA - 2026/09/18
PB - bioRxiv
DO - 10.64898/2026.09.18.752569
UR - https://www.biorxiv.org/content/10.64898/2026.09.18.752569v1
AB - Membrane permeability remains a major challenge in cyclic-peptide drug discovery, where optimization often depends on estimating changes relative to measured reference compounds. Existing predictors mainly model absolute permeability, while heterogeneity across literature sources complicates pooled training and comparative evaluation. We present CycDelta, a multimodal residue-graph model trained directly on within-source permeability differences. CycDelta combines directed message passing with pretrained Uni-Mol monomer embeddings, physicochemical descriptors and assay information in a shared encoder for reference and query peptides. To evaluate this comparative formulation under controlled conditions, we establish CycDeltaBench, comprising in-distribution, scaffold-based out-of-distribution, external and zero-shot unseen-assay tests. On the out-of-distribution task, CycDelta achieved Pearson 0.69 versus 0.46 for the strongest baseline. It also led all evaluated methods on two external data sets and achieved zero-shot Spearman 0.70 versus 0.32 for the strongest baseline on an unseen assay. Residue-level attribution relates predictions to monomer contributions, while gradient-guided screening supports prioritization of permeability-enhancing analogue hypotheses. These results position CycDelta as a transferable and interpretable framework for comparative permeability prediction and cyclic-peptide optimization.
ER -