Friday, April 19, 2013

Combining and conquering at the CAPRI 5th evaluation meeting

After the beer and curry-fest of Manchester and the EGI Community Forum last week, e-ScienceTalk is taking a more leisurely sojourn in a sunny, cobbled Utrecht for the CAPRI 5th Evaluation Meeting. The strapline for the event is ‘Combine and Conquer'. Many processes in the body’s cells are driven by large, complex molecular networks of proteins. The 3D structures of these combined complexes are often the key to their function. But deciphering their detailed structure is difficult, because you need access to large pieces of kit such as X-ray synchrotrons or  nuclear magnetic resonance machines, plus these protein complexes can be fiendishly tricky to crystallise. Lower resolution techniques such as small angle X-ray scattering can give clues and increasingly this data is being used as a starting point for computational modelling. Predicting, modelling and understanding these large complexes is making an important  contribution to drug and protein design – but how confident can you be that the models are correct? CAPRI (Critical Assessment of Predicted Interaction) is an international effort to assess the performance of these methods, by inviting developers to test their algorithms on the same protein targets. This week’s meeting focuses on assessing the performance of docking methods in predicting the 3D structure of complexes and is extending the challenge to predicting binding and de novo interface design.

The opening keynote was by Piet Gros of Utrecht University and winner of the prestigious Dutch Spinoza research prize in 2010. Actually, Gros comes from Dokkum in North Holland, a neat coincidence with the focus on ‘docking’ interactions between proteins at the event. “Although if you think Utrecht is flat, it’s a hill compared to Dokkum,” he remarked.

Gros’s talk highlighted the complement system, which is an ancient part of our immune defence found in the blood. The complement system is formed by around 30 large plasma proteins and cell surface receptors. This system then recognises and eliminates bacteria, viruses and altered host cells, while protecting healthy cells. It links our inbuilt and acquired immunity. “Essentially, it cleans the garbage,” said Gros.

You won’t see a ‘complementology’ department in a hospital, but disruptions in this systems can lead to a range of health problems. It is linked to auto immune conditions such as rheumatoid arthritis, stroke and heart attacks and injuries due to a loss in blood supply, for example to the brain. Genetic changes in complement proteins can lead to kidney conditions and eye problems, such as age related macular degeneration (AMD).  They even have a role in infections, such as EHEC a type of E.coli bacteria, which led to the most expensive drug on the planet being used 2 years ago in Germany during an outbreak.

Using structural studies, Gros’s team has revealed the molecular mechanisms responsible for their functions, such as central amplification, protection of the host body and what happens at the start when a membrane-attack complex forms. In molecular terms, the structural re-arrangements involved are large – about 100 A when a typical atom is a few Angstroms in size (one angstrom is one ten billionth of a metre).

Alexandre Bonvin of the WeNMR project asked Gros to tell the computational modellers in the room what might be lacking from the toolkit he needs for his work. Gros didn’t pinpoint an example, but agreed that the docking modelling and structural work with X-ray crystallography were indeed complementary fields (no pun intended I guess). “In the end we should be able to model everything,” he said, “because otherwise it means we still don’t understand it.”

Over the classic local 'borrel' later, I asked Dr Gros about the computing he uses to support his structural work. Essentially he uses an in-house computing cluster, and does not foresee a need for expanding out to grid computing or the cloud at the moment. This makes an interesting contrast with the WeNMR project community, which relies heavily on grid portals to help them to work on NMR structural data. This presents a challenge for e-Infrastructure providers such as when reaching out to the ‘long tail’ of science – how to engage users with international federated resources when increases in computing power make home grown resources so attractively simple, but maybe not scalable? Does this mean that the long tail has no need of scalable solutions to solve their questions – or can the questions themselves scale up?

No comments: