I am looking for a suitable graph representation where the nodes/vertices are molecules with 2 main variables: structural characteristic and a real number property (for example biological potency). The network is build up by molecules by matching structural properties and it can be rather complex with many clusters of high degree vertices (ie molecules that are structurally similar to each other). However the representation should give a clear picture of potencies, although high detail accuracy is not required: some sort of classification is adequate. I played with JUNG with a graph having edge weights as the change in potency and the ISOM layout thinking that it reserves (as much as possible) the edge weights but with not good results. Something like radial layout could be ideal, as it gives a clear picture of potency distribution, but this obviously would suffer from crowding of vertices that are closer to the center. As, obviously, my knowledge is restricted, is there some layout that you would suggest and is practical for a non expert to implement?