Re: [IMP-dev] Proposal: replace BasicScoreFuncParams with PairPotential

[Daniel Russel <drussel@gmail.com>]

> Maybe some pseudocode
> would elucidate this.
So lets look at evaluating an excluded volume potential for all nearby  
pairs.
So we would have
class NearbyPairsRestraint {
	NearbyPairsRestraint(PairScoreFunction *f);
	evaluate(da) {
		foreach nearby pair:
			f->evaluate(p0,p1, da)
	}
}

Then we use a very general class which applies a ScoreFunc to the  
distance between two spheres:
class SphereDistanceScore {
	SphereDistanceScore(FloatKey radius, ScoreFunc *f);
	evaluate(p0, p1) {
		radius= p0->radius()+p1->radius();
		dist = distance(p0,p1);
		return f->evaluate(dist-radius);
	}
}
sf= IMP.SphereDistnaceScore(FloatKey("radius"), IMP.Harmonic(0, .1))
r= IMP.NearbyPairsRestraint(sf)

If we decided we want to move to a realish Leonard Jones potential,  
then we would replace the SphereDistanceScore since LJ isn't a simple  
function of distance. We could write a little class (ignoring the  
disassociation energy)
class LeonardJonesPairScore {
	evaluate(p0,p1) {
		radius = p0.get_radius()+p1.get_radius();
		dist= distance(p0,p1)
		rat=radius/dist
		return rat^12-rat^6;
	}
}
Then you would just do
r= IMP.NearbyPairsRestraint(particles,  
IMP.LeonardJonesPairScore(IMP.FloatKey("radius")))
Doing this would take a lot of work in the current framework as you  
would have to rewrite all the involved restraints.

Basically, by writing either a trivial PairScore or a trivial  
ScoreFunc you could easily generate complicated custrom restraints. Or  
you could choose whether you want the center distance or the sphere  
distance or the x distance or something else and what potential you  
want and compose things that way and then simply apply them to all  
nearby pairs or all pairs from a list or whatever.