# Prediction of protein-protein interactions using point transformer and spherical Convex Hull graphs

## TL;DR

This paper introduces PT-PPI, a new deep learning framework that uses geometric graphs and point clouds to predict protein-protein interactions more accurately than existing methods.

## Contribution

The novel use of the Spherical Convex Hull method and Point Transformer network for PPI prediction is introduced.

## Key findings

- PT-PPI outperforms existing models on the PINDER benchmark dataset.
- Surface geometry and sequence embeddings complement each other for robust predictions.

## Abstract

Accurate predictions and large-scale identification of protein-protein interactions (PPIs) are crucial for understanding their inherent biological mechanisms and protein functions in virtually all biological processes. Nowadays, graph-based deep learning models have made significant contributions in modeling proteins with physicochemical and geometric features. However, most of these models rely on conventional graph construction methods, such as radial cutoff or k-nearest neighbor (k-NN), which often produce sparse and weakly connected graphs, limiting the ability of neural networks to exploit the spatial relationships between nodes. To address this, we introduce PT-PPI, a geometric deep learning framework that combines protein surface point clouds with geometric graphs. Protein surfaces are encoded as oriented point clouds enriched with geometric features, then transformed into sparse, well-connected graphs using the hyperparameter-free Spherical Convex Hull (SCHull) method. These graphs are processed by a Point Transformer network, with representations coupled to ProstT5 sequence embeddings. Evaluations on the PINDER dataset show that PT-PPI surpasses LLM-based (D-SCRIPT), graph-based (GCN, GAT, Struct2Graph), and hybrid sequence-structural-based models (SpatialPPIv2). Ablation studies confirm the complementary value of surface geometry and sequence information, demonstrating that geometric deep learning on protein surfaces and point cloud representations offers a promising approach that opens the doors for further research on large-scale interactome mapping and the understanding of protein function.

•PT-PPI is a novel geometric deep learning framework for protein-protein interaction (PPI) prediction.•Implements the Spherical Convex Hull (SCHull), a hyperparameter-free method for constructing sparse and connected geometric graphs.•Represents protein surfaces as oriented point clouds enriched with geometric and chemical features.•Leverages a Point Transformer network to capture complex spatial relationships within protein structures.•Achieves state-of-the-art performance, outperforming existing sequence-based, graph-based, and hybrid models on the PINDER benchmark.•Ablation studies confirm the complementary value of surface geometry and sequence embeddings for robust prediction.

PT-PPI is a novel geometric deep learning framework for protein-protein interaction (PPI) prediction.

Implements the Spherical Convex Hull (SCHull), a hyperparameter-free method for constructing sparse and connected geometric graphs.

Represents protein surfaces as oriented point clouds enriched with geometric and chemical features.

Leverages a Point Transformer network to capture complex spatial relationships within protein structures.

Achieves state-of-the-art performance, outperforming existing sequence-based, graph-based, and hybrid models on the PINDER benchmark.

Ablation studies confirm the complementary value of surface geometry and sequence embeddings for robust prediction.