Improving Protein Subcellular Localization Prediction with Structural Prediction & Graph Neural Networks

Geoffroy Dubourg-Felonneau, Arash Abbasi, Eyal Akiva, Lawrence Lee

Keywords: proteomics, protein structure, AlphaFold, Graph Neural Networks, Language Model, ensembling

TL;DR: A lightweight model to ensemble LM ans GNN representations of proteins that beats the state of the art for subcellular localization.

Abstract: We present a method that improves subcellular localization prediction for proteins based on their sequence by leveraging structure prediction and Graph Neural Networks. We demonstrate that Language Models, trained on protein sequences, and Graph Neural Nets, trained on protein's 3D structures, are both efficient approaches. They both learn meaningful, yet different representations of proteins; hence, ensembling them outperforms the reigning state of the art method.