Functional modes of proteins are among the most robust ones

It is shown that a small subset of modes which are likely to be involved in protein functional motions of large amplitude can be determined by retaining the most robust normal modes obtained using different protein models. This result should prove helpful in the context of several applications proposed recently, like for solving difficult molecular replacement problems or for fitting atomic structures into low-resolution electron density maps. Moreover, it may also pave the way for the development of methods allowing to predict such motions accurately. In the case of two-domain proteins, it is well known that a few low-frequency normal modes can provide a fair description of their large amplitude motion upon lig-and binding[1, 2, 3]. More recently, it has been shown that this is also true for proteins with more complex architectures[4, 5, 6], as long as their functional motion is a collective one, i.e. if it concerns large parts of the structure[7, 8, 9]. For instance, a single low-frequency mode of the T form of hemoglobin is enough to describe accurately its conformational change upon oxygen binding[5]. This result has been successfully applied for exploiting fiber diffraction data[10], solving difficult molecular replacement problems[11, 12, 13], or fitting atomic structures into low-resolution electron density maps[13, 14, 15]. The principle of these applications is to perturb a known structure along its low-frequency modes so as to get a deformed structure that is consistent with low-resolution biophysical data, which are obtained after the protein has undergone some large amplitude conforma-tional change. It was also shown that when variations of a few key distances are known, through spectroscopic measurements for instance, it is possible, using linear response theory, to identify which modes are the most involved in the conformational change[16, 17]. However, if such experimental data are missing, it is difficult to guess which low-frequency modes are the functional ones. Hereafter, we show that they are among the most robust ones, i.e. among the most conserved modes when different descriptions of a given protein are considered. The robustness of the functional modes was recognized when it was shown that they can be obtained[7, 8, 9] with simple protein descriptions, like Elastic Network (EN) models[18, 19, 20]. Herein, this property is used so as to identify them. First, standard normal modes were calculated for a set of five proteins of different sizes and architectures after preliminary energy-minimization. The CHARMM program[21] was used, with …