Page - 96 - in Charge Transport in DNA - Insights from Simulations

AParametrizedModel toSimulateCT inDNA Figure7.4: Distributions of IP differences between neighboring bases in a polyA se- quence. Shown are the distributions for the IP difference of nucleobases A2-A1 (left) andA3-A1 (right) obtained fromanatomisticMDsimulation (orange) and fromthemodeled simulation. Finally, it is important tomake sure that the time series of IP is not the same in every simulation. Pilot testswith different randomphase shifts for every internal andenvironmentalfluctuation showedhugefluctuationsof the resulting statistics. Due torandomlygenerated interferencesof thecosine functions,bothmeanvalues and standarddeviations took different values in every conducted simulation. An improved approach is to start the time series of IP at a randomly chosen starting point. For this, a single randomnumber is drawnand set as a phase shift for the entire sumof IP generating cosine functions. This effectively solves the described issueoccurringwhenusingawhole set of randomnumbers. Correlationof theNeighboringNucleobases The IP of neighboring nucleobases are correlated[124]. This is caused by the fact that all nucleobases are influenced by the same environment, and they are physi- cally close toeachother.MDsimulations showthat IPofneighboringbaseshavea correlationcoefficientof0.5up to0.7while the correlationcoefficient is still0.3 for secondneighbors. Tomodel these correlations, a phase shift of the environmental fluctuationswas introduced for the individualnucleobases. Themagnitude of the necessary phase shift can be obtained by comparisonwith the correlation coefficients obtained from MD simulations. Unfortunately, pilot calculations showed that this approachdoesnot lead to reasonablephase shifts. A better parameter tomonitor is the IP difference of the (neighboring) nucleobases. 96