Dear Kay, Thank you for your remark, and sorry - I did not pay attention that Guenter made a copy to CCP4bb. I ment Fig. 7 and corresponding Figure caption in our old article :
Lunin, V.Y., Afonine, P. & Urzhumtsev, A.G. (2002) "Likelihood-based refinement. 1. Irremovable model errors.". Acta Cryst., A58, 270-282. This paper is not specifically devoted to the water analysis but has it as one of examples. The effect is strong since water molecules are placed somehow "systematically" in the unit cell forming a shell following the molecular border. Obviously, the effect of model distortion is diminished/disappears when using stereochemical constraints but nevertheless this means that without these waters X-ray data may push the (surface) atoms somehow in a opposite way, to fit the gap of missed waters, and indeed MAY TRY making the model locally worse. The Maximum Likelihood target (instead of the Least-Squares one) takes care 'statistically' of these missed waters (as well as other missed atoms) and improves the situation but I agree that implicit including of 'riding waters' (similar to 'riding H' atoms) may be useful - obviously, tools and their detailed tests are required. Have a nice Sunday (not at all sunny here today), Sacha ________________________________________ De : Kay Diederichs [kay.diederi...@uni-konstanz.de] Date d'envoi : dimanche 3 novembre 2013 08:28 À : Alexandre OURJOUMTSEV Objet : your email to Günter, and his reply copied to CCP4BB Dear Sacha, Günter seems to have copied his answer to you to CCP4BB, but unfortunately I do not see your original email. Which paper did you refer to when saying "Fig. 7"? thanks, and all the best! Kay
