RosettaCon 2012 GrandChallenges

Rosetta does a a lot of stuff. But what are major challenges facing us over the next 5-10 years? If you were/are a new grad student or postdoc, what big problem should you be trying to crack? (And how in the hell are we going to actually do it?) This year we're going to have a few "Grand Challenge" discussion sessions, where we discuss big picture goals for Rosetta. We'll start with a 5-minute overview from the session chairs, then breakout to discussion in various places. Then reconvene for short talks from the chairs about ideas that were floated.

Where: we have a lot of places to meet. Where each meeting will take place depends on the size of the different groups.

We have:

Chapel theather: 140 seats, theater, projector

Woodpecker: 100 seats, theater, projector

Nuthatch: 30+ seats, round table, projector

Flicker: 30+ seats, round table, projector

Dipper: 30+ seats, round table

Motion and Entropy - chairs Tanja and Oliver

Missing Physics - chairs Phil and Yifan

De novo Design - chairs Sarel and Nobu

If you've come to RosettaCon, chances are that you've read quite a few papers that start with "De novo design of X1 is a stringent test of our understanding of phenomenon X". These papers then demonstrate X1 and end by stating that now X2, X3, etc. will all be a piece of cake and will rid the world of the curse of Z. That's all very nice, but you'd be forgiven if at this point you were scratching your head wondering what you've learned about phenomenon X, be it protein folding, the design principles of active sites, etc. In this session we'll address this question. It turns out that a very exciting aspect of de novo design is that it can teach us what we still don't understand about molecular biology. In this way, de novo design inevitably points the way forward to the next set of challenges and questions that need to be addressed.

We'll review the history of de novo design, demonstrating that de novo design is not a new science as people might think, but has a deep history. Then, we'll review some landmark papers (dwelling very briefly on each) focusing mostly on the unexpected results from these experiments. We'll ask what are remaining challenges in de novo design (so many! but please come up with your own as well), and what can be done to address them.

This will be a discussion and we're hoping for a lot of input from the audience.

Convergence and stopping criteria - chairs Rich, Noah and Charlie

AKA, "When do I stop this crazy thing!?"

We have no consistent stopping or convergence criteria for Rosetta ... For many protocols we have good working procedures to determine when we have sampled enough, but this is often carried out by aux code that is poorly understood and often not used properly by other groups using the code. Is this a huge flaw with the code? Is this lack of convergence criteria for most protocols the source for 95.8 % of global Rosetta miss-use?

This could be a big problem as stopping/convergence criteria: 1) are often based on many thousands of Rosetta runs (as is the case with prediction) and thus would be challenging to implement, 2) these criteria will be different based on small changes to a protocol (designing one site, designing a surface, designing two chains), and 3) principled methods for approaching this problem might just tell us that we are never fully sampled (which would also lead to a global Rosetta misuse pandemic).

There is most certainly no silver bullet that will fix these issues across all protocols, but it might be worth discussing.

Marrying Bench and Server - chairs Rhiju and Jacob

You just finished running your awesome new code and boy does that funnel plot look sweet. But how do you actually test your predictions? Or maybe you finally got that gel to run beautifully, and now you're itching for a structural model of what's going on. But how to actually model it?

Both experiments and simulations are awesome, but the two worlds don't always see eye to eye. How can bench science better inform computation, and vice versa?

• Can what we do actually inform our day-to-day understanding of macromolecular (or even cellular) biology? Can we convince squishy biologists of this?
• Can squishy biology actually inform what we do? How do you iterate between a cellular readout and ideas for modeling or design?
• How can we improve current models for making experiment and computation play nicely?
• What are some radically different ways to merge the two?
• Can one make a general model for interfacing experiment & simulations?
• Should we do different experiments? Or run different calculations?
• Distance restraints work, but they're awfully boring. What are some other crazy things we could do?

Both experimentalists and computationalists are invited!