All current (best) approaches for RNA secondary structure prediction from sequence are based on Turner's model, the best available energy model, but which has limitations. We are studying ways to improve RNA secondary structure prediction, mainly by using a more realistic energy model. We are formulating and studying three different problems: (1) designing a stochastic local search algorithm for secondary structure prediction which could use any energy model; (2) deriving better rules of the model by analysing known structures and by combining two different models from the literature; (3) learning a set of unknown parameters of the model, which reflects more closely our knowledge about secondary structures.

I hope to continue this work for my PhD thesis. I'm planning to defend my proposal in the next few months, so your feedback will be very valuable. I'm going to talk for at most 30 minutes, so that we have at least 30 minutes for discussion/brainstorming.