Page - 90 - in Charge Transport in DNA - Insights from Simulations

AParametrizedModel toSimulateCT inDNA 7.1 Creating theElectronicCouplings MD simulations were performed for DNA sequences containing all 10 possible base-pair steps. The ECwere calculated in everyMD step (1 fs) over a period of 100ps, and the time serieswere analyzed toobtain themagnitudes andcharacter- istic frequenciesof thefluctuations. Figure7.1: EC calculated from MD simulation of 100 ps. Left: Time series of EC shownfor500 fs calculated for5base-pair steps. Right: Power spectrumof the time series from100ps simulation, averaged overfive base-pair steps. Figure7.1 shows the timeseriesof theECinanMDsimulationof500 fs calculated for five base-pair steps in a polyA sequence. These time serieswere analyzed in termsof aFourier transformationover thewhole 100ps simulation time toobtain power spectra. To reduce thenoise, all fivepower spectrawere averaged toobtain the relevantpeakspresent inall base-pair steps. At this point, no characteristic frequencies canbe evaluated from this power spec- trum. Therefore, a straightforward approach (e.g. tomodel the fluctuations as a sumofgoniometric functions)doesnot seemtobe feasible. An alternative approach to parametrize theEC is to create randomnumberswith the correct statistics. To this end, EC time serieswere obtained fromMDsimula- tions of 20nswith the EC calculated every 2ps. Since the sign of the EChas no physicalmeaning, beingdependent on the signof thewave function, theobtained time series contain only absolute values of the EC. This leads to two probability distributions for the EC, onewith a positivemean value and another onewith a negativemeanvalue.Hence, the resultingdistributionof theabsoluteEC is a sum of twoGaussian-type functions. 90