/geometricus_deep_learning

Exploring geometric deep learning algorithms with rotation invariant moments for protein structures

Primary LanguageJupyter Notebook

Exploring geometric deep learning algorithms with rotation invariant moments for protein structures

Done:

  • Graph representations of proteins with 16 3D-rotation invariant moment descriptors per residue, using Geometricus tool [1]
  • Unsupervised graph embeddings using InfoGraph [2] - a contrastive method for graph level representation learning

To be done:

  • Train on AlphaFold 2 structures
  • Explore embeddings
  • Performance test on protein similarity search

References

  1. Janani Durairaj, Mehmet Akdel, Dick de Ridder, Aalt D J van Dijk, Geometricus represents protein structures as shape-mers derived from moment invariants, Bioinformatics, Volume 36, Issue Supplement_2, December 2020, Pages i718–i725, https://doi.org/10.1093/bioinformatics/btaa839
  2. Fan-Yun Sun, Jordan Hoffmann, Vikas Verma and Jian Tang, InfoGraph: Unsupervised and Semi-supervised Graph-Level Representation Learning via Mutual Information Maximization, International Conference on Learning Representations, 2019, https://arxiv.org/abs/1908.01000