We would like to draw your attention to another paper, http://www.biomedcentral.... which was published two months ago on BMC Bioinformatics. That is because both papers used almost the same biophysical model to infer protein binding energies (e.g. chemical potential and nonlinear parameters fitting by minimizing the negative log-likelihood), and all discussed limitation of previous methods such as QPMEME fails to find solution due to an assumption of very low protein concentration. However, interestingly, one paper applied the method on in vitro datasets but the other one was used to analyze in vivo protein-DNA interactions.