[phenixbb] Judging AutoSol Output

Tom Terwilliger terwilliger at lanl.gov
Sun Apr 30 12:42:20 PDT 2006

Hi Angela,

Peter makes several excellent points. Here are a few more notes... The file 
to look at first is "AutoSol_summary.dat"  The final lines in your 
AutoSol.output are the same as that file.  Here are the useful lines and 
how to interpret them

>>Solution # 1   SCORE:110.810151796 Dataset #1   FOM: 0.34 ----------------

110 means 110 sigma above random -> very likely correct

>>Solution 1 using HYSS on ec6046s-peak.sca_ano_1.sca. Dataset #1
>>Dataset number: 1
>>Dataset type: sad
>>Datafiles used: ['ec6046s-peak.sca']
>>Sites: 16 (Already used for Phasing at resol of 2.22)      Refined Sites: 
>>16  Sites in each NCS group: 4  NCS correlation: 0.661541164
>>Experimental phases in: solve_1.mtz_0
>>Density-modified phases in: resolve_1.mtz_0
>>HA sites (PDB format) in: ha_1.pdb_0
>>Model in: resolve_1.pdb
>>   Residues built: 362

Out of 588, this is reasonably good

>>   Side-chains built: 11
>>   Overall model-map correlation: 0.49

Not that great a correlation, but ok at this stage

>>Scaling logfile in: dataset_1_scale.log
>>HYSS logfile in: ec6046s-peak.sca_ano_1.sca_hyss.log
>>Phasing logfile in: solve_1.prt_0
>>Density modification/build logfile in: resolve_1.log_0
>>  Score Type:     Resolve CC    Resolve R     Skew   NCS
>>Raw scores:       0.655        0.492        0.157        0.662
>>Z-scores:        38.788       24.380       14.565       33.077

all Z-scores are very high -> strong solution.  Particularly convincing is 
the NCS correlation of 0.66 and Z-score of 33.

Why didn't solve find all 16 sites?  Solve uses a patterson-based method 
for sites, AutoSol uses direct methods. Could be that in this case the 
patterson method didn't work well.

I hope that helps!
-Tom T

At 04:18 PM 4/24/2006 -0600, agb838 at mail.usask.ca wrote:
>I have some basic questions about the output from Phenix. Basically I
>have a data set that gave a solution, but I am not exactly sure how to
>tell about the quality of the solution from the outputs. The overall
>score is 110. I am not sure exactly what that refers to or how good/bad
>it is. I have attached the AutoSol.output file. Any basic suggestions at
>what I should be looking at, or if this is even the right file, would be
>helpful. The resolve.mtz and resolve.pdb look alright. They are a good
>start, but much improvement is needed. If anyone has some suggestions as
>to how to improve the output, that would also be helpful.
>I have also run the data through solve/resolve through the command
>prompt and it repeatedly locates four heavy atom sites (the data set is
>SAD). Yet, when Phenix is used, 16 sites are found. Why the difference
>and how do I tell what is working better and why.
>Thank you,
>Angela Hoffort
>phenixbb mailing list
>phenixbb at phenix-online.org

Thomas C. Terwilliger
Mail Stop M888
Los Alamos National Laboratory
Los Alamos, NM 87545

Tel:  505-667-0072                 email: terwilliger at LANL.gov
Fax: 505-665-3024                 SOLVE web site: http://solve.lanl.gov
TB Structural Genomics Consortium web site: http://www.doe-mbi.ucla.edu/TB 

More information about the phenixbb mailing list