Unifying evolutionary and thermodynamic information for RNA folding of multiple alignments

Seemann, Stefan E.; Gorodkin, Jan; Backofen, Rolf

There are two main problems for a combined probabilistic model. First, there is no simple way to weight the different information sources in such a combined model. And second, the structure prediction would be based on a maximum likelihood or maximum a posteriori (MAP) approach. Recent work by Carvalho and Lawrence (29) has shown that this approach often does not yield to the desired result. Basically, the implicit assumption of a maximum likelihood or MAP approach is that the structure with the highest probability is also the structure where the ensemble of close neighbors has the highest probability mass, which is often not the case. Hence, Carvalho and Lawrence proposed different new classes of estimators, which included estimators already used in sequence alignment and RNA secondary structure prediction.