view article

Figure 2
(a) Application of the Hamilton R-factor ratio test to comparison of Uoverall and pure TLS models. The structure being refined is a tryptophanyl-tRNA synthetase homolog from Entamoeba histolytica (PDB entry 3hzr). The simple model 1 contains six ADP parameters Uijoverall. The more complex model 2 contains six TLS groups for a total of 120 ADP parameters. The effective weight of the geometric restraints wgeom is unknown, so we calculate the function DF(1)/DF(2) over all possible values of wgeom and show that it is less than the observed R-factor ratio RG(1)/RG(2) = 1.19 everywhere in this range. The quantity wADP is not relevant because neither model contains ADP restraints. (b) Application of the Hamilton R-factor ratio test to comparison of pure TLS and Biso models. In this case the pure TLS model 1 with 120 ADP parameters is the simpler model. The more complex model 2 contains 17 732 Biso parameters, one for each atom. In this comparison both wgeom and wADP are needed but unknown, so DF(1)/DF(2) becomes a two-variable function depending on both. According to the Hamilton ratio test, the more complex model is justifiable only when the R-factor ratio RG(1)/RG(2) = 1.04 (yellow surface) is greater than DF(1)/DF(2) (purple surface). This condition holds only along the far-right edge of the plot corresponding to ADP restraint weights so tight that they approach the limiting condition of a constraint to equal Biso for all atoms.

Journal logoSTRUCTURAL
ISSN: 2059-7983
Follow Acta Cryst. D
Sign up for e-alerts
Follow Acta Cryst. on Twitter
Follow us on facebook
Sign up for RSS feeds