ISSN print edition: 0366-6352 ISSN electronic edition: 1336-9075 Registr. No.: MK SR 9/7 Published monthly

Prediction of apoptosis signal-regulating kinase 1 (ASK1) inhibition with machine learning methods

Zheng-Kun Kuang, Qing Huang, Hui Pan, Xiaoling Duan, and Lixia Huang

Hubei Key Laboratory of Purification and Application of Plant Anti-Cancer Active Ingredients, College of Chemistry and Life Science, Hubei University of Education, Wuhan, China

E-mail: zhengkunkuang@hbut.edu.cn

Received: 10 January 2024  Accepted: 1 May 2024

Abstract:

Apoptosis signal-regulated kinase 1 (ASK1) has recently confirmed as an attractive therapeutic target for drug discovery. The development of small molecule inhibitors of ASK1 has attracted increasing attention. In this work, the rich suite of artificial intelligence methods and descriptor packages available on the OCHEM platform were used to develop regression and classification models for ASK1 inhibitory activity. For regression models, a consensus model was built based on three individual models (LibSVM_Dragon, RFR_QNPR, and LibSVM_RDKIT), which provided the determination coefficient (r²) of 0.69 for the test set. For classification models, the consensus model was developed with ASNN_RDKIT, RFR_Dragon, and RFR_PyDescriptor models and achieved good overall prediction accuracy on test set with 97%. Additionally, we compared the models developed with different algorithms. The results showed that combining various types of descriptors is necessary to improve the predictive ability of the model, and deep learning models provided no advantage over traditional machine learning models in relatively low-data regime. To further explore the relationship between ASK1 inhibitory activity and structural characteristics, several important molecular properties and structural fragments were detected. The results indicated that in silico models could be useful for the virtual screening of ASK1 inhibitors. The important molecular properties and representative structural fragments may be helpful for the discovery of new ASK1 inhibitors.

Keywords: Apoptosis signal-regulated kinase 1; ASK1 inhibitors; Machine learning; In silico models; Virtual screening

Full paper is available at www.springerlink.com.

DOI: 10.1007/s11696-024-03499-y

Chemical Papers 78 (9) 5563–5576 (2024)