Efficient Integration of Molecular Representation and Message-Passing Neural Networks for Predicting Small Molecule Drug-like Properties

Shreyas Bhat Brahmavar,Mrunmay Mohan Shelar,Revanth Harinarthini,Bandaru Hemanth Sai Krishna,Nahush Harihar Kumta,Ojas Wadhwani,Raviprasad Aduri

DOI: https://doi.org/10.26434/chemrxiv-2024-jj94j

2024-03-15

Abstract:The physicochemical properties of a drug molecule determine its metabolism properties. There have been hybrid quantum mechanics approaches with computer-aided drug design and recent supervised machine-learning approaches to predict these properties of small-molecule drugs. However, these methods are low in accuracy and computationally expensive. To get around this problem and improve the performance of a model that predicts the properties of drug molecules, we came up with a novel architecture that uses a "bond order matrix" and structural information to improve molecular graph representations and information in the molecule. Message-passing neural networks (MPNNs) are a framework used to learn local and global features from irregularly formed data invariant to permutations. We take advantage of MPNN architecture and introduce a “semi-master node,” a unique way of representing the functional groups in a small molecule and aggregating features obtained from the functional groups, in anticipation of reverse engineering small molecules given the desired physicochemical properties. This novel architecture and molecule representation were evaluated on the QM9 dataset, which has 133,000 stable small organic molecules with nine heavy atoms (CONF) out of the GDB-17 chemical universe. The metric for evaluating the model's performance is DFT error, an estimated average error of the properties of each molecule. Our models have shown a performance gain of ~10%.

Chemistry

What problem does this paper attempt to address?

The main focus of this paper is how to predict the physical and chemical properties of small molecule drugs more effectively. Current methods are limited in accuracy and computationally expensive. To address this issue, the researchers propose a new architecture that combines "bond sequence matrix" and "semi-major nodes" in a message passing neural network (MPNN). The bond sequence matrix is a molecular representation method that fully retains information about the types and properties of chemical bonds, while the semi-major nodes are used to integrate features of functional groups to reverse engineer small molecules with specific physical and chemical properties. The main contributions of this paper are as follows: 1. Design of an enhanced molecular input representation, namely the bond sequence matrix, to ensure that information about bond properties and types is not lost. 2. Introduction of semi-major nodes in the MPNN architecture to input structural information such as cyclic structures and aromaticity, which have an impact on many physical and chemical properties (such as dipole moments and potential energy). 3. This improved molecular representation method enhances model efficiency and reduces computation time and memory requirements. The paper is evaluated on the QM9 dataset, which is a database containing 133,000 stable small organic molecules with various physical and chemical properties. The new model shows an improvement of approximately 10% in predictive performance based on DFT error as the performance metric. Through this approach, the researchers aim to improve the accuracy and computational efficiency of predicting drug molecule properties, which is crucial for the drug design and development process.

Efficient Integration of Molecular Representation and Message-Passing Neural Networks for Predicting Small Molecule Drug-like Properties

Neural Message Passing for Quantum Chemistry

Dual Message Passing Neural Network for Molecular Property Prediction

An Adaptive Graph Learning Method for Automated Molecular Interactions and Properties Predictions

Message-passing neural networks for high-throughput polymer screening

Ab initio Accuracy Neural Network Potential for Drug-like Molecules

Flexible dual-branched message passing neural network for quantum mechanical property prediction with molecular conformation

Simplicial Message Passing for Chemical Property Prediction

Understanding the Limitations of Deep Models for Molecular Property Prediction: Insights and Solutions.

Proximity Graph Networks: Predicting Ligand Affinity with Message Passing Neural Networks

Flexible Dual-Branched Message-Passing Neural Network for a Molecular Property Prediction

Molecular Property Prediction: A Multilevel Quantum Interactions Modeling Perspective

Integrating Chemical Language and Molecular Graph in Multimodal Fused Deep Learning for Drug Property Prediction

ABT-MPNN: an atom-bond transformer-based message-passing neural network for molecular property prediction

A self-attention based message passing neural network for predicting molecular lipophilicity and aqueous solubility

Multimodal fused deep learning for drug property prediction: Integrating chemical language and molecular graph

Molecular Sharing and Molecular-Specific Representations for Multimodal Molecular Property Prediction

Molecular Mechanics-Driven Graph Neural Network with Multiplex Graph for Molecular Structures

Transferable Multilevel Attention Neural Network for Accurate Prediction of Quantum Chemistry Properties via Multitask Learning

GMPP-NN: a deep learning architecture for graph molecular property prediction

Introducing CARATE: Finally speaking chemistry through learning hidden wave-function representations on graph-based attention and convolutional neural networks