2014 Fiscal Year Research-status Report
Development of an intelligent dynamic docking pipeline for improving molecular docking simulations
| Project/Area Number |
26730152
|
|---|
| Research Institution | Okinawa Institute of Science and Technology Graduate University |
|---|
Principal Investigator |
HSIN KunYi 沖縄科学技術大学院大学, 統合オープンシステムユニット, 研究員 (60604155)
|
|---|
| Project Period (FY) |
2014-04-01 – 2016-03-31
|
| Keywords | Docking Simulation / Molecular Dynamics / Network Pharmacology / Machine Learning / Molecular Interaction / Drug Discovery |
|---|
| Outline of Annual Research Achievements |
In order to precisely and efficiently predict the binding potentials of a given test compound against a large number of targets, we have been developing a screening pipeline which is mainly composed of molecular docking and molecular dynamics (MD) simulations.
For validation, we applied this screening pipeline to PDBbind data set containing about three thousand protein-ligand complexes. The screening pipeline showed a good prediction performance in accessing the ligand binding potentials by evaluating the correlations between the docking scores and the experimental binding affinities. The correlations could be improved from 0.74 to >0.8. We also assessed the accuracy in predicting the selectivity of kinase inhibitors compared with in-vitro competition binding assays, showing a good consistency with the proposed bioassay results. We are now applying MD to refine the docked complexes after docking, subsequently assessing the prediction accuracy.
In order to perform the molecular simulation with ease, we have collected various protein identities (e.g. Gene ID, synonymous/alternative protein names) and have built a “protein identity-to-structure mapping system”. Through it, users can simply give a protein identity to retrieve the corresponding protein structures for simulation, particularly useful for performing high-throughput virtual screening.
Through the collaborations with the Systems Biology Institute (SBI, Tokyo) and The Institute of Medical Science (IMSUT), The University of Tokyo, we are applying our methods to discover anti-influenza agents with well progress.
|
| Current Status of Research Progress |
Current Status of Research Progress
1: Research has progressed more than it was originally planned.
Reason
Based on our previous study which focused on the optimization of scoring function for docking simulation, we could take over the efforts and achievement from it for our current works. Given the advanced hardware support from our university (OIST; 沖縄科学技術大学院大学), we could apply the facility of the HPC (High-performance computing) equipped more than 5,000 cores for computing to our study. The progress of development and validation therefore significantly outperformed initially planned. We are now deploying a powerful GPU workstation for the need of MD simulation. We have initially tested the efficiency of the GPU machine and it shows tremendous performance for running MD. For example, a typical 40,000 atom system will run at a rate of about 40-50 ns per day. To achieve that on a typical CPU cluster, it may need 64 CPUs or cores.
Through the close collaborations with external institutions (SBI and IMSUT), we could have plenty opportunities to validate our methods and to refine our screening protocol accordingly. Experts in systems biology and artificial intelligence (AI) from SBI provide great support in high-quality curated molecular pathway maps (e.g. FluMap and MAPK pathway) required by our screening and the implementation of network pharmacology. Biologists and virologist in IMSUT advise the validation results based on bioassay. Such a unique collaboration can greatly benefit and speed up the progress of this study.
|
| Strategy for Future Research Activity |
We will apply our prediction methods to practical cases. Currently we are collaborating with SBI and IMSUT for the discovery of anti-influenza agents and drug targets using curated pathway map (FluMap) and high-precision molecular docking simulation. Using the method that we are developing, we have discovered some compounds that might have the potentials to interact with proteins of host cell. Through the validation using siRNA screening, those hit compounds were seen bioactivity in-vitro. Our methods therefore show the contribution in reducing the number of tests for follow-up bioassay, and we will apply it to screen larger chemical library.
We will also screen a chemical library derived from herb database (漢方Kampo: traditional Japanese medicine), so that we can discover herb compounds with druggability. Together with a deeply curated molecular pathway map like FluMap, we can accordingly interpret the function of herbs, such as Maoto, by modern medical concepts instead of traditional eastern mysteries.
We are developing an open-type web-based resource equipped with our screening pipeline. A user-friendly GUI interface has also been elaborately designed and will be deployed to the web site, providing efficient methods for molecule preparation, parameter specification and result inspection. Through the web site, drug developers can carry out a network-based screening with ease.
|
| Remarks |
We have been developing a web-based & open-type prediction system for investigating "systems pharmacology" of a given compound freely accessible for the drug discovery community.
|
