Quantitative Biology > Quantitative Methods
[Submitted on 20 Jun 2024]
Title:Geometric Self-Supervised Pretraining on 3D Protein Structures using Subgraphs
View PDFAbstract:Protein representation learning aims to learn informative protein embeddings capable of addressing crucial biological questions, such as protein function prediction. Although sequence-based transformer models have shown promising results by leveraging the vast amount of protein sequence data in a self-supervised way, there is still a gap in applying these methods to 3D protein structures. In this work, we propose a pre-training scheme going beyond trivial masking methods leveraging 3D and hierarchical structures of proteins. We propose a novel self-supervised method to pretrain 3D graph neural networks on 3D protein structures, by predicting the distances between local geometric centroids of protein subgraphs and the global geometric centroid of the protein. The motivation for this method is twofold. First, the relative spatial arrangements and geometric relationships among different regions of a protein are crucial for its function. Moreover, proteins are often organized in a hierarchical manner, where smaller substructures, such as secondary structure elements, assemble into larger domains. By considering subgraphs and their relationships to the global protein structure, the model can learn to reason about these hierarchical levels of organization. We experimentally show that our proposed pertaining strategy leads to significant improvements in the performance of 3D GNNs in various protein classification tasks.
Submission history
From: Michail Chatzianastasis [view email][v1] Thu, 20 Jun 2024 09:34:31 UTC (581 KB)
Current browse context:
q-bio.QM
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
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.