AMIBio is pleased to welcome Sebastian Will (University of Vienna) next Thursday 28/02, 14h30, at LIX in room Philippe Flajolet.

Abstract: Designing (i.e. generating) objects with highly specific properties is a fundamental task with broad applications. For example, for biotechnological applications it is desirable to design sequences of biological macromolecules (RNAs or proteins) such that the molecules adopt specific target structures. This task is generally known as negative design, since the designed sequences must avoid favorable energies for all structures but the one or multiple target structures themselves. In contrast, positive design aims exclusively at favorable energies of the target structures; typically this supports the negative design objectives as well. Infrared abstracts such positive design tasks as the generation of objects with multiple highly specific complex properties. This generalization enables applying general algorithmic methods like the modeling as constraint networks and the targeting of specific feature values by multi-dimensional Boltzmann sampling. This method allows generating distributions that are narrowly centered around targeted feature values, such that satisfying complex properties becomes effective. The method is ultimately based on dynamic programming over tree decompositions that generates Boltzmann distributed samples efficiently with respect to the treewidth due to the targeted features. After outlining the methods in Infrared as well as the modeling based on the Infrared library, I present results from our Infrared-based implementation of positive RNA design “RNARedprint v2”, which can simultaneously target specific energies of multiple RNA structures. Finally, I discuss other applications of the design framework in Infrared, specifically the generation of complex background models.