Pepsi-SAXS: an adaptive method for rapid and accurate computation of small-angle X-ray scattering profiles

Abstract

A new method called Pepsi-SAXS is presented that calculates small-angle X-ray scattering profiles from atomistic models. The method is based on the multipole expansion scheme and is significantly faster compared with other tested methods. In particular, using the Nyquist-Shannon-Kotelnikov sampling theorem, the multipole expansion order is adapted to the size of the model and the resolution of the experimental data. It is argued that by using the adaptive expansion order, this method has the same quadratic dependence on the number of atoms in the model as the Debye-based approach, but with a much smaller prefactor in the computational complexity. The method has been systematically validated on a large set of over 50 models collected from the BioIsis and SASBDB databases. Using a laptop, it was demonstrated that Pepsi-SAXS is about seven, 29 and 36 times faster compared with CRYSOL, FoXS and the three-dimensional Zernike method in SAStbx, respectively, when tested on data from the BioIsis database, and is about five, 21 and 25 times faster compared with CRYSOL, FoXS and SAStbx, respectively, when tested on data from SASBDB. On average, Pepsi-SAXS demonstrates comparable accuracy in terms of χ to CRYSOL and FoXS when tested on BioIsis and SASBDB profiles. Together with a small allowed variation of adjustable parameters, this demonstrates the effectiveness of the method. Pepsi-SAXS is available at http://team.inria.fr/nano-d/software/pepsi-saxs.

Assessment of data‐assisted prediction by inclusion of crosslinking/mass‐spectrometry and small angle X‐ray scattering data in the 12th Critical Assessment of protein Structure Prediction experiment

“…In the case of homo‐oligomeric targets, the Grudinin group started by modeling the complex by first assembling the multimeric structure using SAM, a symmetric assembly protocol based on fast Fourier transform. Their best multimeric models were then relaxed by minimizing the χ values with respect to pre‐computed normal modes …”

“…Their best multimeric models were then relaxed by minimizing the v values with respect to pre-computed normal modes. 47 Other positive results for the top performers include monomeric targets Ts942 and Ts947. For these targets, the Grudinin group improved the quality of their own Ts942 and Ts947 models over their T0 equivalents with DGDTTS of 9.9 and 17.7, respectively.…”

Prediction of protein structure with the coarse‐grained UNRES force field assisted by small X‐ray scattering data and knowledge‐based information

“…The probability distribution P calc ðrÞ should be calculated over the volumes occupied by all atoms of the molecule and the water molecules of at least the first solvation shell. However, it should be noted that the water molecule of the solvation shell should be taken with a smaller weight compared with that of the protein atoms because SAXS measures only the difference between the solvation-shell water and bulk water (see, eg, table 9 in reference 18 ). This is a challenge even when the all-atom representation is used and even a greater challenge at the coarse-grained level.…”

“…In this approach, usually the scattering‐intensity profiles are calculated for each conformation from the Debye formula and the best‐fitting model is subsequently selected or weights of the conformations are fitted to reproduce the experimental scattering profile; some approaches use the SAXS‐derived distance distribution . The scattering profiles are usually calculated by using the CRYSOL program, which also takes into account the solvation shell; alternative approaches have also been developed . Less frequently, SAXS data are included directly, through a penalty term, in the effective energy function in conformational sampling.…”

“…13,16 The scattering profiles are usually calculated by using the CRYSOL program, 17 which also takes into account the solvation shell; alternative approaches have also been developed. 5,7,18 Less frequently, SAXS data are included directly, through a penalty term, in the effective energy function in conformational sampling. The corresponding…”