mode warpNtrace: running ARP/wARP for autobuilding

THIS IS STILL CONSIDERED AN VERSION!

Well, just run arp_warp.sh mode warpNtrace and see what happens.
You probably may have a cup of coffee or a glass of beer while your protein is being autobuilt. Do not get disappointed if the protein does not get traced completely ( AFTER ALL THIS IS AN VERSION).
Look at the connectivity index which is output after every building cycle. If it is higher than 0.4 you have a good chance of improving it to more than 0.9, i.e. tracing almost the whole thing. If it starts much lower, and no usable chains are reported even after the first building cycle, you can start on the O manual ... or think about the next synchrotron visit ...
Here is a typical example (psp, provided within the ARP/wARP  suite) of how the model for a 475 residues protein gets better and better, with a good (i.e. properly measured low resolution) native dataset (kindly provided by Peter Metcalf) extending just to 2.0 Å resolution:

Chains	12	Residues	60	Connectivity Index	0.49
Chains	18	Residues	111	Connectivity Index	0.56
Chains	25	Residues	176	Connectivity Index	0.64
Chains	25	Residues	219	Connectivity Index	0.70
Chains	27	Residues	250	Connectivity Index	0.74
Chains	21	Residues	264	Connectivity Index	0.77
Chains	20	Residues	326	Connectivity Index	0.84
Chains	24	Residues	347	Connectivity Index	0.84
Chains	19	Residues	394	Connectivity Index	0.89
Chains	17	Residues	413	Connectivity Index	0.92
Chains	15	Residues	425	Connectivity Index	0.93
Chains	10	Residues	426	Connectivity Index	0.95
Chains	9	Residues	432	Connectivity Index	0.95
Chains	9	Residues	437	Connectivity Index	0.96
Chains	7	Residues	437	Connectivity Index	0.96
Chains	6	Residues	448	Connectivity Index	0.97
Chains	5	Residues	443	Connectivity Index	0.97
Chains	5	Residues	445	Connectivity Index	0.97
Chains	4	Residues	446	Connectivity Index	0.97

Quite impressive, isn't it? And a good lesson to learn from the above: If you run the default 100 cycles, then you end up with 17 chains and 413 residues. Not bad indeed, but with 100 cycles more you get 446 residues, literally the whole thing, in just 4 continuous chains. That's much, much less work for the graphics. When the connectivity index is just above 0.9, a few extra cycles and CPU time can save you lots of work.
Tricks and Tips

• Look at the 'Connectivity index' after every refinement cycle: If after the first on or two autobuilding cycles it is more than 0.6 then it will almost invariably increase to above 0.9 which means that almost the whole protein will be autobuilt. If it is around 0.45 and 0.6 you have very good chances that it increase, but its usually dependent to data quality. If it is below 0.4, you need either very high resolution native data or better phases and you are most likely (but not necessarily) wasting CPU time.
• This script will first look for a file called warp_free.brk which it will use as a starting model. If this file doesn't exist it will try to build it's own initial model from the FoFOM.ext map (that is created after warp averaging). If this is not there either but you have defined experimental phases it will start by building a free atoms model into the map calculated with phases from that model, refine this model for one cycle as free atoms (i.e. equivalent to skip) and then go on with autobuilding !
• If you have a free atoms model from warp refinement without averaging (i.e. model_1_rc3.brk), copy it to files/warp_free.brk of your working directory.
• If you used warp averaging the script will look for the FoFOM.ext map and go on from there.
• If you have already good phases just run the script, see above what it will do.
• If you start from a molecular replacement model, just copy it to a file called files/warp_free.brk.
• If you are not satisfied with the results from the number of cycles you did, and you think there is additional room for improvement, just copy warpNtrace.brk to warp_free.brk (both in the files/ directory) and restart.
• A very important issue is what you do if you have known heavy atoms in the structure. Usually you know where the heavy atom is, and you want to include it in the refinement. You will have to manually insert the line relevant to that atom in the warp_free.brk file. Then the script will take care not to delete it during autobuilding, etc. The line should look like this:
  ATOM   5578 FE   IUM Y   1       2.322  -0.095   9.294  1.00  5.14  26
  Take care that the IUM notation is used, a valid solvent chain is used (X,Y,Z or W), the element name is 'left justified' and the correct atomic number is given in the last column, else the atom will be ignored. And, yes, the atomic number should correspond to an atom heavier than oxygen. For multiple atoms, simply add multiple lines.
• The script arp_warp_setup.sh sets the number of internal REFMAC cycles depending on which protocol you chose and on the resolution in the MTZ file. For the Heavy protocol, i.e. this application, the number of cycles are 1 for d<1.1 Å, 3 for $1.1\le d < 1.3$ and 5 if the resolution d is between 1.3 and 1.5 Å. These values may well be rubredoxin specific ...If you feel your case requires different numbers, simply edit the warp.par file and change the it rrcyc value. Oh yes, and make sure you know what you are doing.
• Even if you have a descent starting map and good data (above 2.0 Å) it can happen that after getting around 80% of the main chain in a few fragments, (connectivity index 0.8-0.9) you can never reach the so much desired 0.99 and a full trace (this is mainly in folds with long floppy disordered loops). The easiest thing to do is first copy the latest autotracing result ( files/ warpNtrace.brk) to files/warp_free.brk and then run mode warpNtrace conservative, which will disallow wARP to remove main chain atoms and increase convergence.