Neural Embeddings for Protein Graphs

Francesco Ceccarelli,Lorenzo Giusti,Sean B. Holden,Pietro Liò

2023-06-07

Abstract:Proteins perform much of the work in living organisms, and consequently the development of efficient computational methods for protein representation is essential for advancing large-scale biological research. Most current approaches struggle to efficiently integrate the wealth of information contained in the protein sequence and structure. In this paper, we propose a novel framework for embedding protein graphs in geometric vector spaces, by learning an encoder function that preserves the structural distance between protein graphs. Utilizing Graph Neural Networks (GNNs) and Large Language Models (LLMs), the proposed framework generates structure- and sequence-aware protein representations. We demonstrate that our embeddings are successful in the task of comparing protein structures, while providing a significant speed-up compared to traditional approaches based on structural alignment. Our framework achieves remarkable results in the task of protein structure classification; in particular, when compared to other work, the proposed method shows an average F1-Score improvement of 26% on out-of-distribution (OOD) samples and of 32% when tested on samples coming from the same distribution as the training data. Our approach finds applications in areas such as drug prioritization, drug re-purposing, disease sub-type analysis and elsewhere.

Quantitative Methods,Machine Learning

What problem does this paper attempt to address?

The paper aims to address several key issues in protein structure comparison and proposes a new framework for generating protein embeddings. Specifically: 1. **Integrating Sequence and Structure Information**: Most existing methods struggle to efficiently combine protein sequence and structure information. The proposed method generates protein representations that include both structural and sequence information through the use of Graph Neural Networks (GNN) and Large Language Models (LLM). 2. **Fast Similarity Computation**: Traditional structure alignment-based methods take too long to compute on large-scale datasets. The proposed method significantly improves the speed of protein structure similarity computation. 3. **Improved Protein Structure Classification Performance**: In the task of protein structure classification, this method improves the average F1 score by 26% on out-of-distribution (OOD) samples and by 32% on in-distribution samples compared to other existing methods. In summary, the paper aims to develop an efficient and accurate protein representation method that combines both sequence and structure information and performs well in downstream tasks such as protein structure comparison and classification. Additionally, the method has a wide range of applications, including drug prioritization, drug repurposing, and disease subtype analysis.

Neural Embeddings for Protein Graphs

Learning Complete Protein Representation by Deep Coupling of Sequence and Structure

GOProteinGNN: Leveraging Protein Knowledge Graphs for Protein Representation Learning

A Graph Neural Network Approach

GGN-GO: geometric graph networks for predicting protein function by multi-scale structure features

GraphPI: Efficient Protein Inference with Graph Neural Networks

Graph embedding-based novel protein interaction prediction via higher-order graph convolutional network

When Protein Structure Embedding Meets Large Language Models

Accurate Predictions of Molecular Properties of Proteins via Graph Neural Networks and Transfer Learning

Integrating Sequence And Network Information To Enhance Protein-Protein Interaction Prediction Using Graph Convolutional Networks

PersGNN: Applying Topological Data Analysis and Geometric Deep Learning to Structure-Based Protein Function Prediction

Graph-based prediction of Protein-protein interactions with attributed signed graph embedding

Decoding the protein-ligand interactions using parallel graph neural networks

Fast protein structure searching using structure graph embeddings

DeepRank-GNN-esm: a graph neural network for scoring protein–protein models using protein language model

Learning protein sequence embeddings using information from structure

ProtNN: Fast and Accurate Nearest Neighbor Protein Function Prediction based on Graph Embedding in Structural and Topological Space

Protein Representation Learning by Geometric Structure Pretraining

Deep Multi-attribute Graph Representation Learning on Protein Structures

Learning Geometrically Disentangled Representations of Protein Folding Simulations