Abstract

Fragment-based molecular replacement exploits the use of very accurate yet incomplete search models. In the case of the ARCIMBOLDO programs, consistent phase sets produced from the placement and refinement of fragments with Phaser can be combined in order to increase their signal before proceeding to the step of density modification and autotracing with SHELXE. The program ALIXE compares multiple phase sets, evaluating mean phase differences to determine their common origin, and subsequently produces sets of combined phases that group consistent solutions. In this work, its use on different scenarios of very partial molecular-replacement solutions and its performance after the development of a much-optimized set of algorithms are described. The program is available both standalone and integrated within the ARCIMBOLDO programs. ALIXE has been analysed to identify its rate-limiting steps while exploring the best parameterization to improve its performance and make this software efficient enough to work on modest hardware. The algorithm has been parallelized and redesigned to meet the typical landscape of solutions. Analysis of pairwise correlation between the phase sets has also been explored to test whether this would provide additional insight. ALIXE can be used to exhaustively analyse all partial solutions produced or to complement those already selected for expansion, and also to reduce the number of redundant solutions, which is particularly relevant to the case of coiled coils, or to combine partial solutions from different programs. In each case parallelization and optimization to provide speedup makes its use amenable to typical hardware found in crystallography. ARCIMBOLDO_BORGES and ARCIMBOLDO_SHREDDER now call on ALIXE by default.

Highlights

  • Extracting information from vast amounts of data with high levels of error and correlation is critical in many sciences

  • ALIXE builds an average phase set for each set of consistent solutions

  • Combining phase sets derived from partial solutions in fragment phasing increases their information content and is effective in providing a better start for the extension into a full structure through density modification and autotracing

Read more

Summary

Introduction

Extracting information from vast amounts of data with high levels of error and correlation is critical in many sciences. Recent work (Oeffner et al, 2018; McCoy et al, 2017) has shown that the signal for an MR search can be estimated before calculation as the expected LLG for a correctly placed model (eLLG). This value will depend on the accuracy of the model, its size and the resolution of the diffraction data. The main use of our software ALIXE (Millan, Sammito, GarciaFerrer et al, 2015) is to combine information from different partial solutions as an effective way to increase the SNR. ALIXE has been generalized for use with normal computers within all of our programs or as a standalone program and is illustrated with examples

Computing setup
External software
Figures of merit and measures of phase similarity
Test data
Distribution of the software
Results and discussion
ALIXE for ARCIMBOLDO
Accelerating performance: timing benchmarks
Concluding remarks
Funding information
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call