Improved fragment-based protein structure prediction by redesign of search heuristics

Abstract

Difficulty in sampling large and complex conformational spaces remains a key limitation in fragment-based de novo prediction of protein structure. Our previous work has shown that even for small-to-medium-sized proteins, some current methods inadequately sample alternative structures. We have developed two new conformational sampling techniques, one employing a bilevel optimisation framework and the other employing iterated local search. We combine strategies of forced structural perturbation (where some fragment insertions are accepted regardless of their impact on scores) and greedy local optimisation, allowing greater exploration of the available conformational space. Comparisons against the Rosetta Abinitio method indicate that our protocols more frequently generate native-like predictions for many targets, even following the low-resolution phase, using a given set of fragment libraries. By contrasting results across two different fragment sets, we show that our methods are able to better take advantage of high-quality fragments. These improvements can also translate into more reliable identification of near-native structures in a simple clustering-based model selection procedure. We show that when fragment libraries are sufficiently well-constructed, improved breadth of exploration within runs improves prediction accuracy. Our results also suggest that in benchmarking scenarios, a total exclusion of fragments drawn from homologous templates can make performance differences between methods appear less pronounced.

Enhancing fragment-based protein structure prediction by customising fragment cardinality according to local secondary structure

“…The only difference between 'standard' Rosetta predictions and SS-Rosetta is that instead of relying on 25 9-mers and 200 3-mers, which are the default numbers of fragments that have adopted by Rosetta for more than 15 years [72,99,100], SS-Rosetta customises them according to the target. Whilst the whole fragment selection process has not been altered, the number of fragments at specific positions is changed.…”

Reliable Generation of Native-Like Decoys Limits Predictive Ability in Fragment-Based Protein Structure Prediction

“…In a previous study [5], we introduced two new sampling protocols designed to realise more extensive exploration of the available set of conformations within individual runs (a brief description of the methods appears in Section 1.2). We demonstrated that when good fragment sets are used, this improved exploration can translate into more frequent and reliable sampling of near-native basins in the energy landscape for many targets.…”

“…In [5], our experiments using older and newer fragment sets suggested that the fragment sets and choice of sampling algorithm do both play important roles. To address the first of the above questions, it is valuable to investigate the rate at which native-like local minima are accessed during runs of our protocols.…”

“…The remainder of the Introduction contains a survey of alternative acceptance criteria for decoys sampled along trajectories of protein structure prediction methods, as well as a description of our sampling protocols originally introduced in [5], which formed the subject of this study. Having established the effectiveness of our protocols in comparison to the popular Rosetta method [8] in [5], we focused all our experiments presented in this paper on the two methods introduced in [5]. Section 2 then describes the results of our experiments, and Section 3 summarises the main findings and identifies opportunities for further development of our protocols.…”

“…We recapitulate here some key features of our bilevel and ILS protocols (full details in [5]), which are the protocols studied in this paper. Our protocols are implemented as modifications of version 3.4 of Rosetta’s AbinitioRelax application [8].…”