In the 2 sets case then the MST never changes (since it is always a single edge between the two parent particles). All the interesting computation is in the LowestRefinedPairScore, which could be passed to a PairRestraint directly. That is all I meant. In Frank's figure, the ConnectivityRestraint just implements the MST level, the LowestRefinedPairScore implements the first level of selection.? I need the msp functionality, i.e., the restraint should only use the shortest inter-fragment distance for scoring. i thought the ConnectivityRestraint is the appropriate restraint type. if PairRestraint does the same i could also use that - however, i have sometimes >2 fragments involved, so anyways, i'll need the ConnectivityRestraint, i guess.