Project description:
By combining three technological pillars: 1) biophysical computational modeling techniques, 2) advances in hardware, and 3) interactive graphical devices, eDrug will provide a real-time interactive sampling technique for drug design. The idea is to combine computer and human power into the next generation of fully interactive computational tools for molecular modeling. The goal is to merge the human knowledge with state of the art affordable computational resources to create an interactive and (capable of) intuition-guided software solution.
Recent advances in biophysics simulations will provide accurate explorations of the interaction space between drugs and the receptors. These, will run on low cost multi-core platforms, tacking advantage of current future development in computing processors required for the mobile market expansion. Interaction with human expertise will be provided through enhanced graphical solutions, allowing quick comprehension of the results and interaction with the software.
Pilot projects with a world-wide pharmaceutical company, AstraZeneca, will test the prototype developments.
Project Results:
Along the eDrug project we have mainly developed: (i) a human-interactive graphical interface for running PELE in a fast and optimized manner; (ii) speeded-up the software and adapted it for the new generation of heterogeneous processors; (iii) tested the technology on a world-wide pharmaceutical company: AstraZeneca; (iv) successfully transferred the technology to the first spin-off company of Barcelona Supercomputing Center: Nostrum Biodiscovery. Thus, the overall goal of the project has been accomplished: the software developed in eDrug (on top of the one previously developed along the PELE Advanced ERC) will be one of the main assets for the new company, aiming at providing more efficient solutions for computational drug (and biomolecules) design. Specific developments (results) that have allowed this successful outcome include:
1) A new visualization engine enhancing significantly the speed and quality (shading and ambient occlusion) in molecular rendering: “High quality illustrative effects for molecular rendering, Computers & Graphics 54, 113-120 (2016)”
2) Development of novel visualization motifs for analysis (qualitative and quantitative) of non-bonding interactions in protein-ligand complexes (just accepted in SCI Vis, July 2016)
3) Development of a graphical interface (PELE-GUI) combining the new molecular visualization engine and motifs (two first points), and a user-oriented design protocol for: i) preparing simulations; ii) visualizing the results; iii) real-time interaction with the simulations (changing parameters, etc.)
4) Porting the software to use graphing processing units (GPU) acceleration and heterogeneous hardware. In particular, the actual code now includes specific routines for both OpenCL and CUDA, and runs in Montblanc-2, the first supercomputer developed from mobile-based ARM processors.
5) External test of the graphical interface by the medicinal chemistry group in AstraZeneca (Gothenburg, Sweden; group directed by Anders Hogner)
6) Development of a business plan for commercialization. This point has actually been carried out by Nostrum Biodiscovery, the company that has finally transferred the technology, and BSC personnel.
7) Registration of the code (PELE) previous to transfer the technology in May 2016.
All this work has been possible, besides the PoC-eDrug grant, by the involvement of the Barcelona Supercomputing Center, which, after receiving the PoC ERC grant has dedicated numerous efforts (and funds) to contribute to the project. In particular BSC has provided a full PhD student (Pedro Hermosilla) developing the molecular visualization tool, and partial time of our in-house UX specialist, Luz Calvo, in charge of the user-oriented graphical interface design, and has paid all IPR expenses.
