SE Neue Entwicklungen in der RNA Bioinformatik

270151 SE; Peter F. Stadler


Most, or maybe all, RNAs are not "naked" in the cell, but bound by a plethora of protein. Our goal is to understand what this does to RNA structure, to protein binding sites, and to the concentration(s) of (competing) proteins. As background material have a look e.g. at
Interplay between single-stranded binding proteins on RNA secondary structure. Lin YH, Bundschuh R. Phys Rev E Stat Nonlin Soft Matter Phys. 2013 Nov;88(5):052707.
Modeling the interplay of single-stranded binding proteins and nucleic acid secondary structure. Forties RA, Bundschuh R. Bioinformatics. 2010 Jan 1;26(1):61-7.
We will assume that we have a set of binding site sequence together with a preference for a sondary structure context, as derived e.g. with Backofena's graphprot. In essence this yields propensities for the binding of a particular protein to RNA structures as a "knowledge based potential", which we can assume to be related by an unknown affine transformation E = a*P + b, where P is the log-odds score for binding from the data.
On the other hand we know how to compute genome-wide partition functions with RNAplfold on a genome wide scale. With Ronny's new and cool constraint handling machinery we can constrain binding sites to their preferred binding conformation and thus obtain the energy necessary to expose a binding site.

Task 1 Find out how we can get the binding energies for all sites and at least an approximation for the interactions between close-by sites without the (infeasible) computation of all possible occupancy patterns.
Task 2 We can safely assume that we also have RNA expression data. So we also want to model the concentration dependence of binding site occupancy.
Task 3 How can we use our insights to e.g. identify relevant clusters of binding sites? Can we see protein-concentration dependent switches in RNA structure that lead to global effects
Task 4 How does this look like if we have two (or more) proteins with non-overlapping (and overlapping) binding sites. Is there any evidence for cooperative behavior? How would this look like.