Related Experiment Video
Updated: Jun 7, 2025
03:37
Author Spotlight: Impact of Intergenic Interactions on Disease-Identifying Dark Biomarkers
Published on: March 1, 2024
643
MVGNet: Prediction of PI3K Inhibitors Using Multitask Learning and Multiview Frameworks
Yanlei Kang1, Qiwei Xia1, Yunliang Jiang1,2
1Zhejiang Province Key Laboratory of Smart Management & Application of Modern Agricultural Re-sources, School of Information Engineering, Huzhou University, Huzhou 313000, Zhejiang Province,China.
ACS Omega
|November 18, 2024
Summary
A new deep learning model, MVGNet, accurately predicts phosphatidylinositol 3-kinase (PI3K) inhibitor activity. This framework enhances the discovery of novel PI3K-targeted cancer therapies.
Area of Science:
- Biochemistry
- Computational Biology
- Pharmacology
Background:
- Phosphatidylinositol 3-kinase (PI3K) is a critical intracellular enzyme involved in cell signaling, with four main isoforms (PI3Kα, PI3Kβ, PI3Kγ, PI3Kδ).
- PI3K inhibitors are recognized as promising molecularly targeted drugs for cancer therapy due to their antiproliferative and pro-apoptotic effects on tumor cells.
Purpose of the Study:
- To develop and validate a novel multiview deep learning framework (MVGNet) for predicting the inhibitory activity of molecules against the four PI3K isoforms.
- To assess the performance of MVGNet against established machine learning and deep learning models.
Main Methods:
- The MVGNet framework integrates fragment-based pharmacophore information and employs multitask learning to capture inter-task correlations.
- The model was trained and evaluated on its ability to predict inhibitory activity against PI3Kα, PI3Kβ, PI3Kγ, and PI3Kδ.
Main Results:
- MVGNet demonstrated superior performance compared to baseline models, including Random Forest, SVM, XGBoost, GAT, D-MPNN, CMPNN, and KANO.
- The model achieved high average AUC-ROC (0.927 ± 0.006) and AUC-PR (0.980 ± 0.002) values on the test set.
Conclusions:
- The proposed MVGNet framework offers a powerful and accurate approach for predicting PI3K inhibitor activity.
- This study provides valuable insights into the structure-activity relationships of PI3K inhibitors, aiding in the design of novel anti-cancer therapeutics.
