Related Experiment Video
Updated: Jan 15, 2026
10:29
Quantitative Structure-Activity Relationship, Activity Prediction, and Molecular Dynamics of Non-nucleotide Reverse Transcriptase Inhibitors
Published on: May 9, 2025
2.2K
NRIP: A Model for NNRTI-RT Interaction Prediction and Enabling Virtual Screening of Anti-HIV Natural Compounds
Jingxuan Qiu1, Yuxi Zhang1, Mengdie Hu2
1School of Health Science and Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China.
Journal of Chemical Information and Modeling
|October 14, 2025
Summary
A new model, NRIP, predicts drug resistance in human immunodeficiency virus (HIV) reverse transcriptase mutations. This tool aids in identifying effective treatments by analyzing drug-resistant and susceptible interactions.
Area of Science:
- Biochemistry
- Drug Discovery
- Computational Biology
Background:
- Reverse transcriptase (RT) is a key enzyme in HIV replication and a primary target for antiretroviral drugs.
- Mutations in RT frequently lead to drug resistance, necessitating methods to predict these interactions.
- Rapid identification of drug resistance-susceptible relationships is crucial for developing effective HIV therapies.
Purpose of the Study:
- To develop a prediction model, NRIP, for non-nucleoside RT inhibitor (NNRTI) and RT resistant-susceptible interactions.
- To enhance the accuracy of predicting drug resistance by incorporating sequence and spatial structural information.
- To establish a virtual screening pipeline for identifying potential bioactive compounds against drug-resistant HIV.
Main Methods:
- Developed the NRIP model using an extreme gradient boosting (XGBoost) classifier.
- Incorporated novel descriptors combining RT mutation sequence information and residue properties with spatial shell structures.
- Trained and validated the model on 4324 pairs of NNRTIs and RT interactions, including 10-fold cross-validation and an independent testing dataset.
- Constructed a multistep virtual screening pipeline involving structure similarity, drug likeness, and molecular docking.
Main Results:
- The baseline sequence-descriptor model achieved a ROC-AUC of 0.886.
- Incorporating spatial descriptors improved the model's performance, reaching a ROC-AUC of 0.967.
- The final NRIP model demonstrated high accuracy on the independent testing dataset with a ROC-AUC of 0.971 and a PR-AUC of 0.974.
- The virtual screening pipeline showed potential for identifying natural compounds from FOODB against resistant HIV strains.
Conclusions:
- The NRIP model accurately predicts non-nucleoside RT inhibitor and RT resistant-susceptible interactions.
- The integration of sequence and spatial structural descriptors significantly enhances prediction accuracy.
- The developed virtual screening pipeline offers a promising approach for discovering novel anti-HIV compounds, including natural products.
