Why is there a discrepancy between the final fitness score and the calculated fitness score (from individual terms) of a docking performed using the GOLD API?
Under some circumstances when using the GOLD API there may be a discrepancy between the calculated fitness score and the final fitness score displayed in the ligand log file. For example, the ligand log file may contain something like this:
GOLD fitness breakdown of terms:
Fitness = S(hb_ext) + 1.3750*S(vdw_ext) + S(hb_int) + 1.0000*S(vdw_int) + S(cust)
The S(cust) term may have been incorrectly included in the final fitness score more than once. Such a discrepancy could be due to a known issue with the GOLD API. For reasons of efficiency, the fitness is not always recalculated, but sometimes simply copied if no changes have been made to an individual during a previous iteration of the GA. To resolve this issue, GOLD must be prevented from re-adding the custom_score term.