Item

Abstract

Due to the complementarity of the protein design and folding problems, progress on either front has consistently advanced the other. Although both problems remain major challenges, computational protein design has benefited amply from protein structure prediction methods. Likewise, the fields of structure prediction and structural biology have widely adopted techniques from the protein design field. The work I present here aims to put forward new protein design as well as structure determination strategies with the objective of achieving maximum precision. Both strategies capitalise on two posits: the first is that localising the sampling problem allows for exhaustive and finer granularity solution searching, while the second is that accelerated temporal dynamics can allow for directed and accurate exploration of energy landscapes. In the presented protein design projects, the level of precision was evaluated by comparing the coordinates from the experimental structures of the designs to their in silico models. Whereas in the structure determination projects, the precision was evaluated by how well a determined structure ensemble reproduces various experimental observables. Since all of the previous design work utilising conserved supersecondary structures has aimed at constructing repeat proteins from amplifying a single fragment, my first project aims at designing an asymmetric globular (i.e. non-repetitive) fold from two unrelated supersecondary structures. I thereby conceive an interface-driven strategy aiming at constructing a viable intramolecular interface across the participating supersecondary structures. I report the successful design of the target fold that agrees with the experimental NMR structure at atomic precision (backbone RMSD of 0.9 Å), where the designed protein was substantially more stable than its closest natural counterpart. Through the second project I aim to demonstrate the capacity of this interface-driven strategy to tackle the more difficult problem of novel fold design. The computational design of novel folds persists as a profound challenge, as in this case the association between structural and sequence features is absent a priori. This has kept most of the previous design efforts within the known fold space. I accordingly have expanded my interface-design methods, with the goal of achieving efficient sampling at maximum topological control. As a demonstration I conceive and design a novel corrugated protein architecture that does not exist in nature. The resulting NMR and X-ray structures for two different designs agree with the in silico models at atomic precision. On the third project I develop a new generalised method for mapping protein conformational populations from NMR data by unravelling the distribution of states that underlie the experimentally acquired average quantities. The CoMAND method does not only provide a quantitative mapping of the probabilities of the constituent microstates, but is also capable of extracting previously untapped structural information and solving structures de novo from a single NOESY experiment. I further present a detailed protocol that produces highly refined, dynamics-based ensembles without any recourse to heuristics or knowledge-based scoring. Finally, I validate the method’s precision by using the refined ensemble to quantitatively predict NMR observables that are orthogonal to the NOESY data.