Apologies for going slightly further off-topic... Last time I had a free half-day to look into sharp, I noticed that the academic license prohibits reverse-engineering. This seemed to put any comparative testing into a slightly grey area. For example, if I find that sharp does the best job refining sites, but bp3 outputs better phases for a dataset due to different representation of phase probabilities*, I've implicitly constructed a primitive model of how sharp is working. This seems close enough to a first step of reverse-engineering that I was concerned.
Could someone confirm that I'm worrying about things I don't need to here? Pete * Purely hypothetical example. Clemens Vonrhein wrote: > Dear Bill,, > > On Sun, Oct 26, 2008 at 07:18:05PM -0700, William G. Scott wrote: >> I'm actually one of those whom has struggled unsuccessfully to install >> Sharp. I can't even remember why, and I am certain it was entirely my >> fault, but this did get me wondering... > > I guess this is often a more general problem: when we're dealing with > a crystallographic problem that our current set of practices doesn't > deal satisfactorily with and we would like to try out another program, > this is usually also a time of great stress and pressure (competitors, > supervisors, finishing PhD etc). Ideally one would want to install and > learn how to use a new software package when there is a half day or > day time ... seldom happens. > > Without getting into details of installation, usually (> 90%) it is > enough to download our packages and then run "./installSHARP -F". I > might be wrong - but this could be a problem of not getting enough > feedback about failed attempts (time pressure?). So please tell us > about installation problems (or any problems) - hopefully this is a > small price to pay for getting the software for free. If you feel like > sending us positive feedback - even better (especially since > crystallographic software doesn't get citations anymore these days > ... maybe hidden in supplementary material that isn't easily visible > through any search system I know of). > >> mlphare and MIR/MAD phasing seems to me the weak point in the ccp4 >> distribution. (I've often switched over to CNS to calculate phases and >> solvent-flatten them.) >> >> Is there any chance your team might be inspired, persuaded or bribed to >> create a drop-in replacement for mlphare, maybe a "sharp light" for >> boneheads like me? Something that uses the same I/O and interface as ccp4? > > A drop-in replacement would be very tricky: the concepts of SHARP are > quite different (there is no such thing as 'native' or 'derivative') - > but this adds a huge amount of flexibility. > > However, in the next 6.1 release there will be a autoSHARP CCP4i > interface already in the CCP4 distribution (it's already there in the > latest 6.0.99 prerelease or the CVS sources). Although this will still > require the installation of SHARP/autoSHARP itself, at least one could > run autoSHARP without the need of running a httpd. But be aware that > using the CCP4i interface does limit you somewhat - a lot of the > additional tools (regarding density modification, checking/finding > sites via LLG maps, fine-tuning SHARP and viewing maps directly) > aren't easily accessible this way. > > But autoSHARP should deal with all the situations that mlphare etc are > dealing with: SIRAS, SAD, MIRAS, MAD etc. And if you want to do (or > need to do) MAD+native+derivative or TaBr-cluster phasing or > site-specific radiation damage or other more complex phasing > scenarios: SHARP itself is the tool for doing all that. > > Cheers > > Clemens >
