Using Machine Learning to Improve Drug Screening for Beta-Lactamase Inhibitors
Author Information
Author(s): Pitakbut Thanet, Munkert Jennifer, Xi Wenhui, Wei Yanjie, Fuhrmann Gregor
Primary Institution: Friedrich-Alexander-Universität Erlangen-Nürnberg
Hypothesis
Can a machine learning-based QSAR model enhance the effectiveness of consensus docking in screening beta-lactamase inhibitors?
Conclusion
The study shows that a random forest-based QSAR model can effectively improve the identification of beta-lactamase inhibitors compared to traditional consensus docking methods.
Supporting Evidence
- An optimized DOCK6 scoring can maximize the success in identifying active molecules by up to 70%.
- Consensus docking can significantly reduce the false positive rate in determining experimental bioactive molecules compared to the best docking.
- Integrating a random forest-based QSAR model into a virtual screening workflow extends the limited success rate of consensus docking.
- An in-house screening reveals the first-time report of three bio beta-lactamase inhibitors.
Takeaway
This study found a better way to find new medicines that fight bacteria by using smart computer programs to help with the testing.
Methodology
The study involved in vitro screening of 89 compounds from a chemical library, followed by molecular docking using AutoDock Vina and DOCK6, and the development of a random forest-based QSAR model.
Potential Biases
Potential bias may arise from the limited dataset and the reliance on specific docking software.
Limitations
The study was limited by the small number of active compounds identified and the inherent challenges of molecular docking accuracy.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website