In addition, our results also refer to all situations where experimental B-factors were compared with other estimates and measures of protein dynamics. For example, there have been several comparisons between experimental B-factors and fluctuations calculated on elastic lanyard models21,22. In addition, the inverse proportionality between B-factors and the packing density of the neighbors of a given atom has been reported and studied49,50. Finally, many MD studies have directly compared atomic rotations from simulations to experimental B-factors with different results11,24,25. It is possible that, in all these cases, part of the deviation from the experiment is in fact due to experimental models that do not accurately account for the level of true microscopic diversity, as our results suggested. In macromolecular crystallography, the concordance between the observed structural factors and the predicted structural factors (rcryst and rfree) is rarely greater than 20%. This is much larger than the experimental error estimate (Rmerge). The difference between Rcryst and Rmerge is the R-factor gap. There is no such discrepancy in low-molecule crystallography, for which calculated structural factors are generally considered more accurate than experimental measurements. Maybe the true noise level of macromolecular data is higher than expected? Or is the discrepancy caused by imprecise phases that capture refined models in local minima? By generating flexion models simulated with the MLFSOM program and including all possible experimental error sources, we show that both are not the case.
The processing of our simulated data revealed values that, for all crystallographic statistics, with the exception of definitive rcryst and rfree, were not distinguished from those of the actual data. These values fell to 3.8% and 5.5% respectively for the simulated data, indicating that the R-factors elevated in macromolecular crystallography are not experimental errors or phase deformations, but an underlying insufficiency in the models used to explain our observations. The current inability to accurately represent the entire macromolecule, with its flexibility and protein-solvent boundary surface, can be improved through synergies between small-angle X-ray scattering, computational chemistry and crystallography. The exciting implication of our discovery is that macromolecular data has considerable hidden and untapped potential to resolve ambiguities in the true nature of the nanoscale, a task that the second century of crystallography promises to accomplish. a measure of the concordance between the amplitudes of the structural factors calculated from a crystallographic model and those from the initial X-ray diffraction data. The R-factor is calculated during each refinement cycle of the structure with the smallest squares to assess progress. The last R-factor is a measure of model quality. The Q method is not simply an inversion of traditional R-factor analysis to group stimuli or test characteristics, although it does result in such an inversion. Simply demonstrated, factor analysis involves a matrix of basic data (fig. 1) in the social sciences, the correlation of columns of characteristics N on a population of n persons.
This method creates a correlation matrix and a set of factors to evaluate certain characteristics according to their clustering or load. Q analysis (as a factor technique and not as a method) is the personal factoring of characteristic correlations, which identifies statistically correlated groups of human beings (i.e. whole individuals, not characteristics). Given the resemblance between Fstart and Fsim, it is perhaps not surprising that the mere omission of the coordinate model used to calculate Fstart gives remarkably low values of Rcryst and Rfree for refinement against Fsim. In particular, we received starting values from Rcryst/Rfree of 7.37%/7.16%, which evolve after 100 cycles in REFMAC 28 at 6.75%/8.06%.
