Xtricorder strategy in voyager and case 2pan
Hi,
I was just looking at a case that Boaz Shaanan asked about. It seems that Phaser worked to solve this back in 2005 (pre-tNCS), but when he tried rerunning the job now it fails. The failure can be reproduced using the A chain from the deposited structure along with the deposited data (both attached).
The problem is that this is an example of 3-fold tNCS that doesn’t neatly fit into the category of commensurate modulation (unless you allow the modulation to be fairly large). The biggest Patterson peak, at 64% of the origin, is located at 0,0,0.303, and the second biggest, at 26%, is located at 0,0,0.394. If you take the first peak and assume NMOL 3, there should be a peak at 0,0,0.606, which is equivalent to 0,0,0.394 if you consider the inversion operator in Patterson maps (or the translation from molecule C to molecule A instead of from A to C). So I’m guessing that the tNCS order algorithm isn’t considering the inversion operator when testing different orders.
Probably related to this, phenix.voyager.xtricorder comes up with 4 hypotheses for 2-fold tNCS: the identity plus 0,0,0.3; 0,0,0.7; 0,0,0.4 or 0,0,0.6. The first two are equivalent to each other, as are the last two, so this is twice as much work (and twice as much branching downstream) as necessary.
Eventually it would be nice if we could suggest xtricorder as either a replacement or adjunct to xtriage, but it’s not yet very user-friendly in that context. One thing that would help a lot is to have an executive summary at the end of the log output.
Randy