Question: Expected behaviour of scoring function parameter overrides in docking protocol parameter file.

[Beetelbrox]

Description

The user guide states the following:

An RbtNullTransform can be used to send messages to the scoring function to modify
key scoring function parameters in order to increase search efficiency

Existing protocol files in the repo all follow this pattern, where any SF parameter overrides are specified inside RbtNullTransform sections. For example in data/scripts/dock.prm we find the following section:

SECTION SETSLOPE_1
	TRANSFORM           		RbtNullTransform
	WEIGHT@SCORE.RESTR.CAVITY	5.0	# Dock with a high penalty for leaving the cavity
	WEIGHT@SCORE.INTRA.DIHEDRAL	0.1	# Gradually ramp up dihedral weight from 0.1->0.5
	ECUT@SCORE.INTER.VDW		1.0	# Gradually ramp up energy cutoff for switching to quadratic
	USE_4_8@SCORE.INTER.VDW		TRUE	# Start docking with a 4-8 vdW potential
	DA1MAX@SCORE.INTER.POLAR	180.0	# Broader angular dependence
	DA2MAX@SCORE.INTER.POLAR	180.0	# Broader angular dependence
	DR12MAX@SCORE.INTER.POLAR	1.5	# Broader distance range
END_SECTION

Nevertheless, checking the code it seems that the sf parameter overrides can be specified for arbitrary transforms, not only NULL ones. Adding a parameter override to an arbitrary transform seems to work fine, for example the following is a valid transform section:

SECTION RANDOM_POP
        TRANSFORM                       RbtRandPopTransform
	WEIGHT@SCORE.RESTR.CAVITY		3.0	# Dock with a high penalty for leaving the cavity
        POP_SIZE                        50
	SCALE_CHROM_LENGTH		TRUE
END_SECTION

I have not found any reference to this in the user guide or example in the docs.

Question

Is the ability to specify parameter overrides in arbitrary transforms intended behaviour, or should it be restricted to the null transform? It seems that the parameter changes are applied before the execution of the transformation itself so I'd say it makes sense to allow it, but wanted to confirm