One of my docking2 docking_local_refine jobs (#17790) output many structures with Fnat=NaN. I am used to seeing Fnat near 1 when the rms value is low and Fnat near 0 when the rms value is high. What would make Fnat give a value of NaN? Should I trust the other scores given for this run? How can I prevent future runs from giving Fnat=NaN?
Fnat is the fraction of native contacts. If for some reason you don't have any native contacts, then you'll get a NaN (not a number - one way to get this is by dividing by zero). If this is the case, you will also likely see the Irms (interface rmsd) also be NaN, for similar reasons (no interface residues, resulting in a divide by zero).
Actually, in docking2 docking_local_refine job 17790, the 20 decoys with the most negative I_sc values have total_score=-106.625 to -107.714, rms=9.699 to 17.262, Fnat=NaN, I_sc=-0.037 to -0.81, and Irms=0.322 to 2.164. Meanwhile, in docking2 local_docking job 17791 (based on the same initial pdb file as job 17790), the 20 decoys with the most negative I_sc values have total_score=-105.021 to -110.599, rms=9.915 to 26.312, Fnat=NaN, I_sc=-2.858 to -3.549, and Irms=0.731 to 6.706.
Does this information change what you think causes the Fnat=NaN results?