Is there a database where I can find an affinity estimate if I provide a given antibody and a given antigen sequence ?
Input : antibody + antigen sequence Output : quantitative binding/affinity estimate
Biology Stack Exchange is a question and answer site for biology researchers, academics, and students. It only takes a minute to sign up.Sign up to join this community
The binding constant of an antibody-antigen complex remains an empirical measurement.
There have been many efforts to solve this problem quantitatively and qualitatively over the years since the original crystal structures of an lysozyme-Fab complex there are over 400 structures at the www.rcsb.org in a search for antibody antigen complexes.
At this point the general methodology is to model the structure of the antibody-antigen complex. Then from the complex a free energy of binding is estimated, or more reliably, variants which might bind more tightly can be modelled.
There are tools available to do this work, but they are still far from perfect. This 2012 paper compares a new antibody-antigen crystal structure against the results of several modeling suites. In Table 4 the authors show that the invariant parts of the structure as well as some of the hypervariable loops are really precisely modelled, but hypervariable region 3 was way off in this case (rms error of about 2 Angstroms). Which is not precise enough to calculate actual free energy of binding.
Currently while free energies of binding and therefore estimates of binding constants are not reliably calculated from protein complex structures. Such software is in practice used to rather compute the relative binding strength of two models (see this 2012 example).
This 2010 paper which describes SnugDock, a newer suite of software, outlines a state of the art antibody-antigen design process (emphasis added).
One could envision a complete computational antibody engineering pipeline starting from the antibody sequence and ending with accurate predictions for optimized antibody-antigen interactions. In this paper we have been successful in reaching the second step by computational docking using computational models of the monomer antibody. The next stages may be additionally challenging. High-resolution complex structures might next be used for computational alanine scanning, computational affinity maturation or alteration of binding specificity. For antibody therapeutics, structures will help define drug mechanisms for regulatory approval, enable epitope mapping and humanize constructs. These applications require varying amounts of resolution and further testing will reveal the full utility of the SnugDock predictions.
As you can see, the method recommended is to create a series of models to try to create tighter relative binding.