Extended Data Fig. 1: All-atom simulations of DNA rotation using the TIP4P-D water model.

a, Rotation angle versus simulation time of a DNA duplex under 100 (red) and 10 (black) mV/nm electric field carried our using the TIP4P-D model of water. The simulation system was identical to the periodic TIP3P simulation system, Fig. 2a. The average rotation rate determined via a linear regression fit is −5.1 and −1.1 degrees/ns at 100 and 10 mV/nm, respectively. We attribute the non-linear scaling to a large statistical error of the 10 mV/nm simulation. For reference, the average rotation velocity observed in our TIP3P simulations under a 100 mV/nm field was −12.8 degrees/ns, Fig. 1c. b, Distribution of the effective torque values. The instantaneous torque values were sampled every 2.4 ps and averaged using 5 ns blocks. The vertical solid lines depict the mean values of the distributions: −0.74 and −0.11 pN nm for 100 and 10 mV/nm. For reference, the average magnitude of the effective torque measured using the TIP3P model of water is about 0.68 pN nm at electric field magnitude of 100 mV/nm, Fig. 2d. Thus, the torque values are insensitive to the water model used, indicating that the shear stress on DNA from the fluid does not depend on the fluid viscosity, as expected. c, Ionic current through a cubic volume of 1 M KCl electrolyte, 5.8 nm on each side, under a 10 mV/nm electric field obtained using the TIP3P, TIP4P-D and out custom implicit solvent models, averaged over 5-ns blocks. For reference, the expected experimental ionic current value is plotted as a dashed line, computed using the experimental 11.0 S/m conductivity of 1 M KCl. The bulk conductivity values computed from the currents are 16.2, 7.2, and 13.5 S/m for the TIP3P, TIP4P-D and our custom implicit solvent models, respectively. The lower than experimental conductivity of the TIP4P-D electrolyte suggests that the model may systematically underestimate the electro-kinetic effects because of the lower than expected electrophoretic mobility of ions.