\documentclass[a4paper, 11pt]{article} \begin{document} \section{Introduction} This document describes the use cases and API's for the program Mosflm. Mosflm is an unusual case, because there are number of significant functions it performs - each of these will have to be a well identified use case. \subsection{Functionality} The program mosflm has the following key pieces of functionality which need to be exposed \emph{via} the API: \begin{itemize} \item{Autoindexing of diffraction patterns.} \item{Estimation of mosaic spread.} \item{Refinement of unit cell.} \item{Integration of diffraction patterns.} \item{Calculation of data collection strategies (deprecated.)} \end{itemize} \subsection{Input} Each of these tasks requires a certain amount of common input - it can only make sense to standardise on the API for this. Of particular importance are the content of the image header and the frames to use in processing. \section{Rules} \subsection{Introduction} There are a couple of rules which are useful when running mosflm, and these should be encoded in the \emph{wrapper} rather than the rest of the expert system because they are specific to this program. These rules pertain to: \begin{itemize} \item{Indexing errors, in particular lattice assignment.} \end{itemize} \subsection{Indexing Errors \& Lattice Assignment} Prototype data set: 1VR9 native (12847) from the JCSG. This is correctly C2 symmetry, but is pseudo-I222. Autoindexing with mosflm, even with the ``refine'' keyword assigned, gives a lattice of I222. However, when this is set the mosaic spread estimation \& refinement fails, which is a pointer that there is something wrong. The estimation fails saying ``you'll have to guess'' or something, and is then refined to a negative value in cell refinement. This should be interpreted as a pointer that the lattice has too high a symmetry. Asserting a spacegroup of C2 corrects this, so clearly this is something important to build into the wrapper. \section{Making Sure Nothing is Lost} FIXED 16/AUG/06 the distortion \& raster parameters decided on in the cell refinement stages need to be recycled to use in integration. This is demonstrated by running an interactive (through the GUI) mosflm autoindex and refine job, then dumping the runit script. The important information is in the following records: { \small \begin{verbatim} Final optimised raster parameters: 15 17 12 5 6 => RASTER keyword Separation parameters updated to 0.71mm in X and 0.71mm in Y => SEPARATION keyword XCEN YCEN XTOFRA XTOFD YSCALE TILT TWIST 108.97 105.31 0.9980 149.71 0.9984 -13 -46 => BEAM, DISTANCE, DISTORTION keywords (note that the numbers are on the next line here) \end{verbatim} } This should make the resulting integration more effective. The idea for this implementation is that the numbers end up in the "integrate set parameter" dictionary and are therefore recycled, in the same way that the GAIN currently works. Note well - some of this conflits with the definition of the FrameProcessor, in which case the relevant information should be passed back to that interface. Have implemented an application xia2process, which should test out this wrapper. However, there seem to be some issues with the implementation when running with scripts, or Labelit as the indexer implementation, which I will need to fix. This is reasonably effective... below are some results illustrating the changes for the 1VPJ/12287 dataset. Note that this is not quite a fair comparison, because the original set had the bad areas of the detector masked out. Still, it's interesting. The order is INFL orig, INFL new, LREM orig, LREM new. { \small \begin{verbatim} Summary data for Project: demo Crystal: 12287 Dataset: infl Overall InnerShell OuterShell Low resolution limit 52.56 52.56 1.74 High resolution limit 1.65 5.22 1.65 Rmerge 0.056 0.034 0.348 Rmeas (within I+/I-) 0.073 0.044 0.465 Rmeas (all I+ & I-) 0.096 0.074 0.522 Rpim (within I+/I-) 0.045 0.027 0.305 Rpim (all I+ & I-) 0.045 0.036 0.283 Fractional partial bias -0.010 -0.032 0.008 Total number of observations 110206 3541 9776 Total number unique 26070 970 3281 Mean((I)/sd(I)) 12.8 24.9 2.4 Completeness 97.9 99.1 87.8 Multiplicity 4.2 3.7 3.0 Anomalous completeness 96.8 81.2 Anomalous multiplicity 2.3 1.7 DelAnom correlation between half-sets 0.583 0.122 Mid-Slope of Anom Normal Probability 1.573 Summary data for Project: demo Crystal: 12287 Dataset: infl Overall InnerShell OuterShell Low resolution limit 52.56 52.56 1.74 High resolution limit 1.65 5.22 1.65 Rmerge 0.055 0.032 0.347 Rmeas (within I+/I-) 0.071 0.041 0.463 Rmeas (all I+ & I-) 0.094 0.072 0.519 Rpim (within I+/I-) 0.044 0.026 0.304 Rpim (all I+ & I-) 0.044 0.035 0.281 Fractional partial bias -0.010 -0.030 -0.038 Total number of observations 111491 3732 9910 Total number unique 26064 979 3304 Mean((I)/sd(I)) 13.2 26.5 2.5 Completeness 97.9 99.3 88.1 Multiplicity 4.3 3.8 3.0 Anomalous completeness 96.9 81.5 Anomalous multiplicity 2.3 1.7 DelAnom correlation between half-sets 0.597 0.166 Mid-Slope of Anom Normal Probability 1.611 Summary data for Project: demo Crystal: 12287 Dataset: lrem Overall InnerShell OuterShell Low resolution limit 52.56 52.56 1.74 High resolution limit 1.65 5.22 1.65 Rmerge 0.052 0.031 0.313 Rmeas (within I+/I-) 0.067 0.040 0.422 Rmeas (all I+ & I-) 0.065 0.038 0.457 Rpim (within I+/I-) 0.042 0.025 0.281 Rpim (all I+ & I-) 0.031 0.020 0.258 Fractional partial bias -0.006 -0.017 -0.071 Total number of observations 106594 3567 8233 Total number unique 25692 975 2969 Mean((I)/sd(I)) 13.3 25.9 2.5 Completeness 96.4 99.1 79.4 Multiplicity 4.1 3.7 2.8 Anomalous completeness 95.1 74.3 Anomalous multiplicity 2.3 1.5 DelAnom correlation between half-sets -0.192 -0.033 Mid-Slope of Anom Normal Probability 0.868 Summary data for Project: demo Crystal: 12287 Dataset: lrem Overall InnerShell OuterShell Low resolution limit 52.56 52.56 1.74 High resolution limit 1.65 5.22 1.65 Rmerge 0.052 0.031 0.313 Rmeas (within I+/I-) 0.067 0.040 0.422 Rmeas (all I+ & I-) 0.065 0.039 0.458 Rpim (within I+/I-) 0.042 0.025 0.281 Rpim (all I+ & I-) 0.031 0.020 0.259 Fractional partial bias -0.007 -0.020 -0.068 Total number of observations 107779 3737 8294 Total number unique 25678 982 2976 Mean((I)/sd(I)) 13.6 27.1 2.6 Completeness 96.4 99.4 79.6 Multiplicity 4.2 3.8 2.8 Anomalous completeness 95.1 74.5 Anomalous multiplicity 2.3 1.5 DelAnom correlation between half-sets -0.215 -0.020 Mid-Slope of Anom Normal Probability 0.862 \end{verbatim} } Note the very slight improvements in $R$, $I/\sigma$ etc. \section{Thoughts} Could get some useful information by parsing the SUMMARY file with the CCP4 loggraph parser - this would give all of the information which is displayed in loggraph, which could save some parsing of the standard output. \end{document}