The scale factors also should be checked for reasonableness. Correct scaling is important! Bad scale factors cause errors in every structure factor, and every point in your map. The quality of the scale factors depends upon the resolution of the diffraction data. Sometimes, as the resolution drops, the scaling parameters go bad earlier and sometimes later.
You have to judge the parameters based on their values and to do that you need a basic understanding of the meaning of the scaling parameters. TNT uses two scale factor parameters not used by other packages. These parameters provide a simple model of the scattering of the bulk solvent. All four of TNT parameters are described below.
`B' is the number which must be added to each atom in the model to get the average B to match the observed structure factors. If B comes out the be 3 it is probably okay. If it is 27 it means that all the B's in your model have to be increased by 27, putting the overall average in the 40's. If you believe that to be true, fine. If not you have to set B to a reasonable value, usually zero. At low resolutions (3Å or lower) the R-value is very insensitive to this parameter. It could assume almost any value and the R-value will come out the same. In these cases B should always be set to zero.
`Ksol' is the ratio of the average solvent density to the average protein density, electron density that is. Protein, being what it is, and water, being what it is, the ratio should be about 0.7. If your mother liquor is denser Ksol will be larger. It is a rare mother liquor that is denser than protein so Ksol shouldn't be larger than 1.0. It certainly should never be negative. Where there are no high resolution data the precision required of Ksol becomes less. 0.7, 0.8, 0.9, 1.0 are all about the same. When the data are lacking and the program can't come up with a reasonable number you simply set it to a value of your choice.
`Bsol' is the average temperature factor of the bulk
solvent. It should be large and positive. It seems to hover
in the neighborhood of 100 to 300. Sometimes it will come out
to be very large ( 
1000). This usually occurs when you have
not given enough low resolution data. The program chooses a
solution which pushes all the solvent scattering into the
unobserved portion of reciprocal space. The way to avoid this
problem is to include the low resolution data.
Sometimes Bsol is very small. In most cases this is because there is no model for the ordered solvent in your molecule and the program is trying to model the scattering from these atoms with the disordered solvent parameters. In this situation you have to force Bsol to be about 150 or so to keep it from interfering with these ordered atoms.
`K' is the conversion factor to take the observed F's from whatever units they are measured in to electrons per unit cell, the units of the calculated F's. Its value cannot be predicted nor can you simply look at it and decided it correctness. The only check you have on it is that if the other parameters are correct it has to be too. However if K is larger than 10 it means that you have lost one digit of precision of your observed structure factors in the formatted HKL file. This means that you should multiply all your data by whatever factor is required to preserve the precision of your original data file before you write it to HKL format.
During refinement you want the individual temperature factors of the model to absorb the signal which causes the parameter B to be non-zero. The model cannot adjust to this signal if the signal has already been eliminated. B must be fixed with a value of zero in all refinement. This requirement is built into the tnt command.
If the B factor offset is too large (> 5) normal TNT refinement will not be able to adjust the individual B's sufficiently for the error to be corrected. When this occurs it is best to use the script $tntbin/add_b to add the value of the B scale factor to each individual B factor in the model. To make this correction type ``$tntbin/add_b init.cor 5.36", with your value of B substituted. The corrected model with be written to add_b_init.cor. Rename this file to init.cor and proceed with refinement.
Additional restrictions must be placed on the scale factors if the data sets do not cover large areas of reciprocal space. The inner resolution limit should be set to the lowest resolution possible, say 20Å, for a good solvent model to be achieved. To properly determine all the scaling function parameters the outer resolution limit should be better than 2Å. This second criterion is often more difficult to achieve. If your crystals do not diffract to better than 2.5Å resolution you will have to fix B at zero even when calculating maps. This is done by adding a ``SET B 0.0'' statement to your control file.
If you do not have data better than 5Å you will have to fix Ksol at a reasonable number (say 0.8). This is done by including ``SET B 0.0 KSOL 0.8''.
If you are working with just a thin shell of reciprocal space you should set all parameters to zero except K. This is done with ``SET B 0.0 KSOL 0.0 BSOL 0.0''.