> Now my query is, whether one should pick water molecules at this low > resolutions or it is totally unscientific to do so?
Your question is justified in intent, but ill phrased. The question you are faced with is “How plausible would the assignment of a given electron density reconstruction feature as a water molecule be?” The answer depends on observational evidence and chemical plausibility. You have the most knowledge about your protein complex and should have some knowledge about chemical plausibility of your proposal. (a) A few questions to consider re. evidence: What is the noise level in your map? How do normal 2Fo-Fc densities compare to difference densities? Density shape? Other isolated mystery density of same levels somewhere? If your maps are excellent and low noise it is not impossible to see a very well bound water molecule at 4A. (b) Plausibility based on prior expectations: Was Mg in the cocktail? Being isoelectronic with HOH (and a favorite companion of DNA in crystallization), it might be a plausible candidate. Anything else heavier, perhaps? SO4, PO4? Perhaps any clues from anomalous data/ano diff maps? Fragments of PEGs? What does the refinement tell you? How did you refine? Does your protocol match the low resolution of the data? Even at the low resolution, do bond length and coordination support a discrete moiety? Distances, geometry, B-factors? If everything points in your favor, you can justify the proposition of a discrete moiety. Your scientific credibility depends on how well your proposition is supported by reasoning from (a) and (b) – probably with heavy emphasis on (b) as you are poorly determined – and not whether you are ultimately right or not. Although I doubt that a water molecule without biological relevance assigned to it has any effect on global refinement stats nor on your career, you can always invoke the rule of parsimony for your model – no explanation is better than an unsupported one. “I don’t know” is a perfectly scientific answer. LGBR
