analiza aiqbench
This research investigates how explicitly encoding molecular scaffolds influences the geometry of self-supervised learning embeddings, finding that such supervision reliably organizes chemical space by structural similarity. While Lorentzian contrastive objectives showed stronger organizational effects than Euclidean ones, neither metric provided a consistent overall advantage across all tasks. The study concludes that leveraging structural cores to shape embedding spaces is beneficial for property prediction, though the specific gains remain dependent on the target task.
Structural Hierarchy and Geometry in Molecular Representation Learning
arXiv:2608.29886v1 Announce Type: new Abstract: Molecular self-supervised learning uses chemical structures to guide which molecular embeddings should be similar. We study whether explicitly encoding a molecule's Bemis-Murcko scaffold and using it to supervise the molecular embedding changes what the model learns. We further test whether this effect depends on the embedding geometry by comparing Euclidean and Lorentz contrastive objectives. Across two augmentation strengths, scaffold-supervised models consistently organize molecules according to both identical and structurally related scaffolds. The resulting embeddings also improve molecular property prediction on several tasks, while the exact gains depend on the predicted property. The effect of scaffold supervision on molecular organization is stronger under Lorentz objectives, but neither geometry provides a consistent overall advantage. These results show that explicitly teaching the relation between a molecule and its structural core can reliably shape the organization of molecular embedding space, while the extent of usefulness of this organization remains task dependent.
Abstract:Molecular self-supervised learning uses chemical structures to guide which molecular embeddings should be similar. We study whether explicitly encoding a molecule's Bemis-Murcko scaffold and using it to supervise the molecular embedding changes what the model learns. We further test whether this effect depends on the embedding geometry by comparing Euclidean and Lorentz contrastive objectives. Across two augmentation strengths, scaffold-supervised models consistently organize molecules according to both identical and structurally related scaffolds. The resulting embeddings also improve molecular property prediction on several tasks, while the exact gains depend on the predicted property. The effect of scaffold supervision on molecular organization is stronger under Lorentz objectives, but neither geometry provides a consistent overall advantage. These results show that explicitly teaching the relation between a molecule and its structural core can reliably shape the organization of molecular embedding space, while the extent of usefulness of this organization remains task dependent.
Machine Learning (cs.LG); Biomolecules (q-bio.BM)
arXiv:2608.29886 [cs.LG]
(or arXiv:2608.29886v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2608.29886 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 243 cuvinte
Subiecte conexe
PIRAMID: Progress and Plans
LessWrong · 27 iul. 2026
Geometry-Consistent Endoscopic Representations for Image-Guided Navigation via Structured Foundation Model Adaptation
arXiv · 17 iun. 2026
Beyond Predictive Accuracy: A Reliability-Aware Audit of Molecular Representations for Human Olfaction
arXiv · 29 iul. 2026
Riemannian Flow Models with Reinforcement Learning for Molecular Crystal Structure Prediction
arXiv · 1 oct. 2026