Comparison of Different Reweighting Approaches for the Calculation of Conformational Variability of Macromolecules from Molecular Simulations

Abstract

Conformational variability and heterogeneity are crucial determinants of the function of biological macromolecules. The possibility of accessing this information experimentally suffers from severe under‐determination of the problem, since there are a few experimental observables to be accounted for by a (potentially) infinite number of available conformational states. Several computational methods have been proposed over the years in order to circumvent this theoretically insurmountable obstacle. A large share of these strategies is based on reweighting an initial conformational ensemble which arises from, for example, molecular simulations of different qualities and levels of theory. In this work, we compare the outcome of three reweighting approaches based on radically different views of the conformational heterogeneity problem, namely Maximum Entropy, Maximum Parsimony and Maximum Occurrence, and we do so using the same experimental data. In this comparison we find both expected as well as unexpected similarities.

Conformational Space Sampled by Domain Reorientation of Linear Diubiquitin Reflected in Its Binding Mode for Target Proteins

“…Thus, overall, the sampled conformational spaces proposed by the two approaches are consistent with each other. Previous studies have reported the high consistency among results from the MaxOcc approach and the maximum parsimony approach SES (sparsest ensemble selection) for some biomolecular systems, which range from marginally flexible systems, e. g., HIV‐1 TAR RNA, to highly flexible systems, e. g., calmodulin [32] …”

“…Previous studies have reported the high consistency among results from the MaxOcc approach and the maximum parsimony approach SES (sparsest ensemble selection) for some biomolecular systems, which range from marginally flexible systems, e. g., HIV-1 TAR RNA, to highly flexible systems, e. g., calmodulin. [32] Second, MaxOcc values were computed based on the paramagnetic data exclusively from the 15 N-distal unit of Ub 2 D39Cp-T1-Tm 3 + and Ub 2 G47Cp-T1-Tm 3 + , which exhibited a significant reduction of MaxOcc around the center to the lower left quadrant of the frame, as shown in Figure 5b. This pattern could simply reflect the fact that the PCSs obtained from the 15 N-distal unit were substantially smaller (À 0.162 to À 0.031 ppm) than that of the 15 N-proximal unit (À 0.289 to 0.568 ppm).…”

Exploring the impact of alignment media on RDC analysis of phosphorus-containing compounds: a molecular docking approach

“…The latter approach requires a large amount of anisotropic data, which are not always available. Another approach, usually used for macromolecules, employs the so-called reweight models [26][27][28][29] (maximum entropy, maximum parsimony, maximum allowed probability, etc.). However, these methods are computationally demanding and fail for more flexible systems.…”

Refinement of molecular dynamics ensembles using experimental data and flexible forward models

“…This makes it possible to either improve the employed force fields [5][6][7][8][9][10][11][12][13] or to directly refine the generated ensembles. In particular, ensembles can be constructed by enforcing agreement with the experiment on the fly 9,[14][15][16][17] or by refining the ensembles a posteriori 18 , using either selection 19,20 or reweighting 16,[21][22][23][24][25] approaches. Methods based on the idea of minimally perturbing the initially generated ensemble are particularly appealing as they maximally use the microscopic information generated by the MD simulation and only modify it when deviation with respect to the experiment is observed 21,26 .…”