Abstract:Protein-protein interactions (PPIs) play a crucial role in numerous biochemical and biological processes. Although several structure-based molecular generative models have been developed, PPI interfaces and compounds targeting PPIs exhibit distinct physicochemical properties compared to traditional binding pockets and small-molecule drugs. As a result, generating compounds that effectively target PPIs, particularly by considering PPI complexes or interface hotspot residues, remains a significant challenge. In this work, we constructed a comprehensive dataset of PPI interfaces with active and inactive compound pairs. Based on this, we propose a novel molecular generative framework tailored to PPI interfaces, named GENiPPI. Our evaluation demonstrates that GENiPPI captures the implicit relationships between the PPI interfaces and the active molecules, and can generate novel compounds that target these interfaces. Moreover, GENiPPI can generate structurally diverse novel compounds with limited PPI interface modulators. To the best of our knowledge, this is the first exploration of a structure-based molecular generative model focused on PPI interfaces, which could facilitate the design of PPI modulators. The PPI interface-based molecular generative model enriches the existing landscape of structure-based (pocket/interface) molecular generative model.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to develop a molecular generation framework that can effectively generate compounds targeting the protein - protein interaction (PPI) interface. Specifically, the researchers noted that traditional structure - based molecular generation models face challenges when dealing with PPI interfaces because there are significant physicochemical property differences between PPI interfaces, traditional binding pockets, and small - molecule drugs. Therefore, designing a compound generation method that can take into account PPI complexes or interface hot - spot residues is an important scientific problem. ### Research Background and Problem Description 1. **Importance of PPI**: - Protein - protein interactions (PPIs) play a crucial role in numerous biochemical and biological processes. - PPI modulators can expand the drug target space and have great potential in drug discovery. 2. **Limitations of Existing Methods**: - Traditional structure - based rational design plays an important role in identifying lead compounds, but PPI targets and traditional drug targets have significant differences in biochemical characteristics (see Table 1). - Different structural features lead to differences in the physicochemical properties and drug - likeness of PPI modulators compared to conventional drugs. - Molecular generation models targeting PPI structures or interfaces are rarely reported in the literature. 3. **Specific Problems**: - How to generate novel compounds that can effectively target the PPI interface? - How to use deep - learning techniques to capture the implicit relationships between PPI interfaces and active molecules and generate compounds with diversity and novelty? ### Solution: GENiPPI Framework To solve the above problems, the researchers proposed a new molecular generation framework - GENiPPI (Interface - aware molecular generative framework for protein - protein interaction modulators). This framework is implemented through the following steps: 1. **Dataset Construction**: - A comprehensive dataset containing PPI interfaces and active/inactive compound pairs was constructed. 2. **Model Architecture**: - **GAT Module**: Graph Attention Networks (GATs) are used to capture the atomic - level interaction features of the protein complex interface. - **CNN Module**: Convolutional Neural Networks (CNNs) are employed to encode the representation of compounds in voxel and electron - density spaces. - **cWGAN Module**: The Conditional Wasserstein Generative Adversarial Network (cWGAN) is used to integrate these features and train to generate compound representations targeting the PPI interface. - **LSTM Decoder**: Long Short - Term Memory (LSTM) is used to decode the molecular embeddings into SMILES strings. 3. **Performance Evaluation**: - The effectiveness of the generated compounds is evaluated by multiple metrics, including QED, QEPPI, and Fsp3. - The performance of GENiPPI is compared with other generation models (such as LatentGAN and ORGAN) in terms of novelty, diversity, and effectiveness. ### Conclusion The GENiPPI framework demonstrates its superior performance in generating new compounds with drug - likeness and PPI - targeting properties, especially in few - shot generation tasks. This framework provides a powerful tool for structure - based PPI modulator design and enriches the existing field of structure - based molecular generation models. --- If you have more questions or need further explanation, please feel free to let me know!

Interface-aware molecular generative framework for protein-protein interaction modulators

De novo molecular design with deep molecular generative models for PPI inhibitors

mPPI: a database extension to visv-ilize structural interactome in a one-to-many manner

Target-specific design of drug-like PPI inhibitors via hot-spot-guided generative deep learning

3D molecular generative framework for interaction-guided drug design

Computational Methods-Guided Design of Modulators Targeting Protein-Protein Interactions (Ppis)

Identification of potential PIM-2 inhibitors via ligand-based generative models, molecular docking and molecular dynamics simulations

In-Pocket 3D Graphs Enhance Ligand-Target Compatibility in Generative Small-Molecule Creation

CProMG: controllable protein-oriented molecule generation with desired binding affinity and drug-like properties

ResGen is a pocket-aware 3D molecular generation model based on parallel multiscale modelling

Computational Modeling as a Tool to Investigate PPI: From Drug Design to Tissue Engineering

A Multimodal Deep Learning Framework for Predicting PPI-Modulator Interactions.

Structure-Based de Novo Molecular Generator Combined with Artificial Intelligence and Docking Simulations

PocketFlow is a data-and-knowledge-driven structure-based molecular generative model

Computational Probing Protein-Protein Interactions Targeting Small Molecules.

Benchmark Study Based on 2P2I(DB) to Gain Insights into the Discovery of Small-Molecule PPI Inhibitors

Computational Approaches for the Design of Modulators Targeting Protein-Protein Interactions

LS-MolGen: Ligand-and-Structure Dual-Driven Deep Reinforcement Learning for Target-Specific Molecular Generation Improves Binding Affinity and Novelty.

Where to next in the evolution of in silico design of modulators targeting protein–protein interactions?

moPPIt: De Novo Generation of Motif-Specific Binders with Protein Language Models

A Simple Way to Incorporate Target Structural Information in Molecular Generative Models