Start with a COMPLETE NATIVE DATASET or you are hurting yourself.

Soak crystals in solution where they can live from few hours to few days. 
Include dissolved heavy atom salts (0.1 to 20 mM) in this soaking solution. 
If you cryo-protect your crystals for freezing, include the same concentration 
of heavy atom in all subsequent cryo-solutions otherwise you will 'back-soak' 
the metal out of your crystal.



Screen many heavy metal species.
Ignore metals that crack crystals at all concentrations.
Improve / optimize soaks that are derivatives.


Rapid screening while data is collecting:
   You want to know if you should continue collecting after analyzing only 10-20° of data. 
   If you don’t see positive signs before 40% (max) of the data are collected, this will not 
   be a derivative with significant phasing power. Also, you need to compare many reflections, 
   so an acceptable first derivative should diffract to at least 4 Å.

	Assay	Amt of data	Software	Acceptable criteria
1	Cell constants/isomorphism	1-2 frames	DENZO	delta a,b,c < 0.6%
2	Intensity chi2 statistics vs. native	3°	SCALEPACK	6 < chi2 < 40 at low res
3	Scaling to native F’s	10°+	SCALEIT (ccp4)	10 < Rscale < 30
4	Normal Probability Statistics	10°+	SCALEIT (ccp4)	Slope>4 Intercept>0.8
5	Difference Patterson (iso/ano)	15-30°	FFT (ccp4)/CNS	Harker peaks > 8 sigma
6	Difference Fourier (iso/ano)	15-30°	FFT (ccp4)/CNS	Atom peaks > 5 sigma

If a derivative does not pass tests 1-4, do not waste any more time collecting data, 
you will not find peaks in tests 5 and 6.