At the end of a long day at the CAPRI meeting, the sessions closed with a brainstorming discussion on the present and future of CAPRI. Which targets should be in the frame for the next phase? Before the discussions got underway though there were a series of lightning talks from early stage researchers – great to see them brought into the spotlight. A session on docking methodology and servers followed. Paul Bates of Cancer Research UK focused on the SwarmDock server, a webservice for generating 3D structures of protein-protein complexes, by looking for the lowest energy solutions. He described the swarming process as a bit like wandering around Utrecht after a few beers – you might end up in the canal, get lost down a cobbled street or “be run over by one of these high speed cyclists.” There was a big laugh of recognition from the delegates for that one.
Anisah Ghoorah from the University of Lorraine, INRIA, pointed out that, “Docking is difficult, but that also makes it interesting.” Ab initio techniques i.e. starting from scratch are too difficult, so there’s a need to use templates and knowledge based approaches. “CAPRI should focus on the difficult cases – leave Google for the simple ones,” he said.
ClusPro was the first fully automated, web based program for computational docking of protein structures. It has been around for 10 years, and has about 3000 users. Sandor Vajda of Boston University had been looking at usage of the server over the course of the year. “First of all you can see that dockers don’t work on the weekend – and they go on holiday in August,” he pointed out. “And no one cares about Christmas!”
Martin Zacharias of TU Munchen wound up the session by presenting protein-protein docking and refinement with a coarse-grained protein model. The computing needed seems very efficient. “It’s possible to do 10,000 docking simulations on a single core in an hour,” said Zacharias. “These happen much faster in reality than they do in silico.”
Finally, the community took at look at some new areas proposed for CAPRI. These include interface predictions, useful for cancer studies and interface water predictions, which are potentially very challenging and are currently underrepresented. Also floated were protein-peptide interactions, protein-oligosaccharides, protein-RNA and even protein-DNA (with a question mark). For affinity challenges, it’s a question of what they both can and want to evaluate. From the discussion, it was clear that they are looking for a mix of easy and difficult targets, as well as having lots of targets to choose from. However, cooperation from the structural community is essential – they need to submit their structures to compare with the predictions. “They won’t solve the targets just to be nice to us!” it was pointed out. So it’s not just protein-protein interactions that are important in the end, but those between the structural biologists and modellers as well.
Anisah Ghoorah from the University of Lorraine, INRIA, pointed out that, “Docking is difficult, but that also makes it interesting.” Ab initio techniques i.e. starting from scratch are too difficult, so there’s a need to use templates and knowledge based approaches. “CAPRI should focus on the difficult cases – leave Google for the simple ones,” he said.
Martin Zacharias of TU Munchen wound up the session by presenting protein-protein docking and refinement with a coarse-grained protein model. The computing needed seems very efficient. “It’s possible to do 10,000 docking simulations on a single core in an hour,” said Zacharias. “These happen much faster in reality than they do in silico.”
Finally, the community took at look at some new areas proposed for CAPRI. These include interface predictions, useful for cancer studies and interface water predictions, which are potentially very challenging and are currently underrepresented. Also floated were protein-peptide interactions, protein-oligosaccharides, protein-RNA and even protein-DNA (with a question mark). For affinity challenges, it’s a question of what they both can and want to evaluate. From the discussion, it was clear that they are looking for a mix of easy and difficult targets, as well as having lots of targets to choose from. However, cooperation from the structural community is essential – they need to submit their structures to compare with the predictions. “They won’t solve the targets just to be nice to us!” it was pointed out. So it’s not just protein-protein interactions that are important in the end, but those between the structural biologists and modellers as well.
Posts
No comments:
Post a Comment