Comment on "Length scale dependence of DNA mechanical properties".

Recent experimental data [1–7] indicate that the elastic wormlike rod (WLR) model of DNA that works well on long length scales may break down on shorter scales relevant to biology. According to Noy and Golestanian (N&G) [8] molecular dynamics (MD) simulations predict DNA rigidity close to experimental data and confirm one scenario of such breakdown, namely, that for lengths of a few helical turns, DNA dynamics exhibit long-range bending and stretching correlations. Earlier studies using similar forcefields [9–13] concluded that (i) MD systematically overestimate the DNA rigidity, and (ii) no deviations from the WLR model are detectable [14, 15]. Here it is argued that the data analysis in N&G was incorrect and that the earlier conclusions are valid. Measuring DNA rigidity by MD requires rigorous analysis of statistical errors. For high accuracy, trajectories must be several orders of magnitude longer than the corresponding relaxation times, and optimal conditions are met for dynamics of one helical turn [12, 13, 16, 17]. Longer fragments are difficult to study because the relaxation times for twisting and bending grow with the DNA length as L and L, respectively [10].