I would like to compare the affinity between 2 difference structures. InterfaceAnalyzer can calculate binding energy with more suitable by including dSASA term. In the description dG_separated/dSASAx100: Separated binding energy per unit interface area * 100 to make units fit in score file. Scaling by dSASA controls for large interfaces having more energy. The factor of 100 is to allow standard 2.45 notation isntead of something like 2.45E-2. When look in the detail, dSASA is calculated from the difference between binding complex and separated complex. So, the larger interface, the greater dSASA, right?. Therefore, the dG should be higher(get closer to zero) when divided by large interface or high dsasa score. It does not make sense to me . Please anyone give me the right explaination.
You're correct in that a larger interface would have a greater dSASA. But it would also (on average) have a greater binding energy, as there are more residue interactions which can make favorable contacts.
So what you tend to see when you're doing binding studies and looking just at dG values is that there are two types of complexes with good binding energies. You get small, highly favorable interfaces where all the interactions are good and complementary, and you also get large, diffuse interfaces where the interactions are so-so, but there's a lot of them because the interface is so large.
That's what dG_separated/dSASAx100 is trying to discriminate between. It's not sufficient to have a high binding energy, but you also want to have a high binding energy *density*. In a lot of complexes it's better to have a smaller interface with a slightly worse Rosetta score as long as the interactions that you are making are better, stronger interactions. You can take it too far, though - you don't want a tiny interface with a single perfect interaction. You need to balance multiple factors - you want a good binding energy, but you also want that to be from a good binding energy density. How to balance those is a bit tricky - I'd recommend looking at a range of know interfaces of native protein complexes similar to your system of interest and seeing how they perform by the various metrics.
Quite tricky . Thanks you very much