After running autoSHARP on 3isy, you have a reasonably complete model built by arp/wARP, 3isy_eden_flat_40.4pc_warpNtrace.pdb, and a solvent-flattened MTZ file 3isy_eden_flat_40.4pc.mtz.

The next step is to use autoBUSTER to refine the model. In this case, we'll refine the model and perform automatic water insertion:

refine -p 3isy_eden_flat_40.4pc_warpNtrace.pdb \
       -m 3isy_eden_flat_40.4pc.mtz \
       -autoncs -L -d initial | tee initial.lis

-autoncs is redundant in this case since there's only one chain, but it does no harm, and it's worth getting into the habit of using -autoncs at all times. -L is the command for using water insertion.

The run starts with Rfact=0.218 Rfree=0.265, and ends (half an hour later on a fast PC) at Rfact=0.190 Rfree=0.234. Looking at the difference maps indicates that there are large difference peaks on the methionines, unsurprisingly since they are mostly selenomethionine. The B-correlation term used by BUSTER prevents the B-factor of the sulfur from being adjusted downwards to account for the higher electron density at the selenium, whilst in the original arp/wARP output you can see that the SD B factor is about half that on the CG and CE atoms.

There are only two methionines, so change them by hand in a text editor. Remember to change the atomic symbol at the end, and the SD atom to be called SE with the SE in the first two columns (IE the E aligned with the C of CG or CE, and the S one left from that) - the PDB format is at times forgiving and at times unbending about the alignment of atom names, so it's worth being unreasonably careful.

refine -p 3isy_withMSE.pdb \
       -m 3isy_eden_flat_40.4pc.mtz \
       -autoncs -L -d initialMSE | tee initialMSE.lis

starts with 247/262 (the Rfact has increased substantially, because we've moved quite a lot of electrons around in what was a reasonably refined input model) and ends with 186/228 - this goes to show that single large scatterers can have a distressingly big impact on Rfree.

Unfortunately, this is so beautifully clear a model that there's almost nothing to build; you will have to use a different example for practice in the use of coot. There might be a residue or two to add at the C-terminus, and maybe some waters are slightly dubious.