*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
A new version of pirate (version 0.4.9) available today at
http://www.ysbl.york.ac.uk/~cowtan/pirate/pirate.html
as linux and OSX/PPC binaries.
This is a substantial reworking, including:
- A significantly simpler user interface. You don't have to specify a
reference structure unless you particularly want to (like the new
buccaneer GUI).
- A complete rewrite of the NCS finding code. It now works a good
proportion of the time, and doesn't take forever (seconds for up to 20
heavy atoms to hours for up to 200 atoms).
- A rewrite of the NCS averaging code. NCS is not applied with the
rest of the phase improvement, but as a preprocessing step. This is a
very conservative approach, and doesn't necessarily exploit all of the
NCS information, but it does avoid bias.
- Rewritten content determination code. This isn't my last word on the
subject, but the new code is robust and fast.
- Automatic weighting of likelihood information from both non-NCS and
NCS sources, independently, using the log-likelihood gain. This corrects
a signfiicant problem with earlier versions, and also provides some
(limited) scope for pirate to cope with badly estimated phase errors
from older programs. (It is still better to use a modern phasing
program, however).
This is the first version I would be confident of including in an
automated pipeline.
There are some rough edges - memory usage has gone up, and some internal
refactoring will speed up the NCS code too. But it should be a
signficant improvement on previous versions.
Kevin
