A cocrystal prediction method of graph neural networks based on molecular spatial information and global attention

Kang, Yanlei; Chen, Jiahui; Hu, Xiurong; Jiang, Yunliang; Li, Zhong

doi:10.1039/d3ce00535f

Cited by 3 publications

(2 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this work, every API-coformer pair was presented to the network twice, once per every separate CNN, as previously suggested [11,27] . This strategy allows artificially increasing the number of inputs available for model training.…”

Section: Deepcocrystal Architecturementioning

confidence: 99%

See 1 more Smart Citation

Deep Supramolecular Language Processing for Co-crystal Prediction

Birolo,

Özçelik,

Aramini

et al. 2024

Preprint

View full text Add to dashboard Cite

Approximately 40% of marketed drugs exhibit suboptimal pharmacokinetic profiles. Co-crystallization, where pairs of molecules form a multicomponent crystal, constitutes a promising strategy to enhance physicochemical properties without compromising the pharmacological activity. However, finding promising co-crystal pairs is resource-intensive, due to the vast number of possible combinations. We present DeepCocrystal, a novel deep learning approach designed to predict co-crystal formation by processing the 'chemical language' from a supramolecular vantage point. Rigorous validation of DeepCocrystal showed a balanced accuracy of 78% in realistic scenarios, outperforming existing models. By leveraging the properties of molecular string representations, DeepCocrystal can also estimate the uncertainty of its predictions. We harness this capability in a challenging prospective study, and successfully discovered two novel co-crystal of diflunisal, an anti-inflammatory drug. This study underscores the potential of deep learning -- and in particular of chemical language processing -- to accelerate co-crystallization, and ultimately drug development, in both academic and industrial contexts.

show abstract

Section: Deepcocrystal Architecturementioning

confidence: 99%

“…Machine learning -which extracts relevant information from chemical datasets [6] -can aid in prioritizing APIcoformer pairs for co-crystallization [7][8][9][10][11] . Current methods, however, might struggle to generalize to previously unseen molecules [12] .…”

Section: Introductionmentioning

confidence: 99%

Deep Supramolecular Language Processing for Co-crystal Prediction

Birolo,

Özçelik,

Aramini

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

Noncovalent Derivatization

Warner

2024

Reference Module in Chemistry, Molecular Sciences and Chemical Engineering

View full text Add to dashboard Cite

Cocrystal Prediction Tool (CCPT): A Web Server for Deep Learning-Assisted Cocrystal Screening and Density Evaluation

Guo,

Yang,

Wang

et al. 2024

Crystal Growth & Design

View full text Add to dashboard Cite

Cocrystallization, as a molecular modification strategy, has exhibited great success in the chemistry and material fields. To reduce time and labor costs in experiments, machine learning of artificial intelligence (AI) has been introduced to accelerate cocrystal development. Despite the theoretical success of the exploits, their usages suffer from professional programmed operation and substantial human intervention, thus hindering their application in practice. To fill up the gap between theory and practical application, we explore a one-step and user-friendly cocrystal prediction web server (CCPT), which integrates two state-of-the-art deep learning models with high generalization and accuracy, including cocrystal screening and density evaluation. Users only need to upload molecule pair files and select the job type. CCPT can automatically perform feature descriptor calculation and prediction tasks (cocrystallization formation and their densities). All prediction results will be packaged into a file format that can be visualized and downloaded on the web page by users. The entire process under the ergonomic graphical interface is code-free operation and a streamlined workflow with minimum human intervention and thus very simple and convenient for the end-user, especially for experimental investigators. CCPT is free and accessible at http://www.scuccpt.cn. We expect that it will be a useful design tool for diverse cocrystal fields.

show abstract

A cocrystal prediction method of graph neural networks based on molecular spatial information and global attention

Abstract: This paper presents a cocrystal prediction model based on molecular point cloud information and a graph attention network (GAT). We firstly expand our experimental dataset for the training purpose. This...

Cited by 3 publications

References 38 publications

Deep Supramolecular Language Processing for Co-crystal Prediction

Deep Supramolecular Language Processing for Co-crystal Prediction

Noncovalent Derivatization

Cocrystal Prediction Tool (CCPT): A Web Server for Deep Learning-Assisted Cocrystal Screening and Density Evaluation

Contact Info

Product

Resources

About