HisF/HisH interface; and hydrogen-bonding between the Ω-loop
and the conserved 49-PGVG sequence (i.e., the oxyanion strand)
adjacent to the HisH active site. The CNA outcome showed to be
of particular help for further manipulation of the allosteric regulation in IGPS, allowing rational design of allosteric inhibitors that
could interfere with the suggested allosteric pathways [32] and
promoting experimental mutagenesis studies [53] that granted
knockout of IGPS allosteric signal propagation. Moreover, the
CNA proved to be a transferable approach that we have successfully
employed to other allosteric systems [24, 25], in conjunction with
other graph approaches involving the eigenvector centrality metric
to account for long-range correlated motions [26] and to dynamical perturbation networks based on inter-residues physical contacts
along MD simulations [54].
5 Notes
1. The CNA method is based on the outcome of standard MD
simulations, and the outcome is strictly related to the simulation time and the window the user decides to analyze. Generally, time windows around 50–150 ns are reasonable, while this
depends on the system size: larger protein usually requires
longer MD simulation time to sample motions possibly relevant to allostery.
2. To get the best possible statistical analysis of correlated
motions, once a time window is selected (e.g., 100 ns), several
“running” time windows of that length should be extracted
from the MD trajectories. As mentioned above, running more
than one independent simulation is strongly suggested. The
correlation coefficients computed for each of these time windows (and for multiple trajectories) can then be averaged out to
provide a single correlation matrix for each system. A preliminary investigation of the CNA outcome as function of the time
window length chosen (e.g., comparing results with 50, 100,
150 ns time windows) is also suggested.
3. For very large proteic systems, the computation of the correlation coefficients might be quite demanding. To overcome this
limitation, a suggested solution is to use linearized MI coefficients, as suggested by Lange et al. and implemented in the
“g_correlation” code [44] of the GROMACS software [38].
4. The code for the CNA method is available under request to the
authors of this chapter. The code provides as output pictures
like those in Fig. 4b and text files that can be readily used to
148
Ivan Rivalta and Victor S. Batista
and the conserved 49-PGVG sequence (i.e., the oxyanion strand)
adjacent to the HisH active site. The CNA outcome showed to be
of particular help for further manipulation of the allosteric regulation in IGPS, allowing rational design of allosteric inhibitors that
could interfere with the suggested allosteric pathways [32] and
promoting experimental mutagenesis studies [53] that granted
knockout of IGPS allosteric signal propagation. Moreover, the
CNA proved to be a transferable approach that we have successfully
employed to other allosteric systems [24, 25], in conjunction with
other graph approaches involving the eigenvector centrality metric
to account for long-range correlated motions [26] and to dynamical perturbation networks based on inter-residues physical contacts
along MD simulations [54].
5 Notes
1. The CNA method is based on the outcome of standard MD
simulations, and the outcome is strictly related to the simulation time and the window the user decides to analyze. Generally, time windows around 50–150 ns are reasonable, while this
depends on the system size: larger protein usually requires
longer MD simulation time to sample motions possibly relevant to allostery.
2. To get the best possible statistical analysis of correlated
motions, once a time window is selected (e.g., 100 ns), several
“running” time windows of that length should be extracted
from the MD trajectories. As mentioned above, running more
than one independent simulation is strongly suggested. The
correlation coefficients computed for each of these time windows (and for multiple trajectories) can then be averaged out to
provide a single correlation matrix for each system. A preliminary investigation of the CNA outcome as function of the time
window length chosen (e.g., comparing results with 50, 100,
150 ns time windows) is also suggested.
3. For very large proteic systems, the computation of the correlation coefficients might be quite demanding. To overcome this
limitation, a suggested solution is to use linearized MI coefficients, as suggested by Lange et al. and implemented in the
“g_correlation” code [44] of the GROMACS software [38].
4. The code for the CNA method is available under request to the
authors of this chapter. The code provides as output pictures
like those in Fig. 4b and text files that can be readily used to
148
Ivan Rivalta and Victor S. Batista