Hi Graeme, There was a CCP4BB thread about this quite recently (14th Nov 2013). I've coped below responses from Edward Berry and Matthew Franklin.
SCALA & AIMLESS have no sigma cutoffs, but TRUNCATE does. According to the documentation, reflections with intensities less than minus 4 standard deviations are rejected. However, in the code this seems to be less than minus 3.7 standard deviations (rather than 4). So for data that has been processed by TRUNCATE, I think that the observed criterion sigma(I) is -3.7. This is hard-wired in the code. It is interesting (perhaps) that this number only seems to be requested for PDB depositions processed by RCSB, PDBe do not seem to ask for this (at least, not the last time I deposited). Andrew Edward Berry: As I understand it this refers to the decision whether an observation is valid or not, and the default value in HKL suite is -3 sigma (note the negative sign). The denzo/scalepack manual explains that while it is important not to exclude observations that are slightly negative due to random errors of measurement, anything that comes in below -3 sigma is likely to be a fluke and should be discarded. I'm not sure whether this refers to measurements before adding partials, or to the summed full reflection observation. anyway, I always put -3s for that value and haven't had any negative feedback from the annotators. Matthew Franklin: HKL2000 (Denzo/Scalepack) use I greater than -3 sigma (that's NEGATIVE 3) as the observed criterion, so that's what you would put down for this entry. There is another place where you're asked to provide an observed criterion for F's used during refinement. I always put down 0 (i.e. use all F's) for this one. I have no idea what Scala does. On 25 Nov 2013, at 09:21, Graeme Winter <[email protected]> wrote: > Hi Folks, > > A xia2 user wrote in asking where to find > > 'observed criterion sigma(F)' and 'observed criterion sigma(I)' > > in the xia2 logs (i.e. from Scala or Aimless or XSCALE)... I have no idea > what they are so will struggle to give a helpful answer ;o) and surprisingly > google was not a lot of use coming up with > > Data processing information : high and low resolution limits, observed > criterion sigma (F) cut-off or observed criterion sigma (I) cut-off, number > of unique measured reflections (all and observed), percent of possible > reflections observed, R-merge I (observed) or R-sym I (observed), details > about the highest resolution shell > > Can anyone point me in the right direction? > > Thanks in advance & best wishes, Graeme
