Figure 1

Fitness landscapes and adaptive evolution for a single binding site (a) Crater landscape (3) and (b) Mesa landscape (4), as a function of the Hamming distance r from the target sequence (within the approximation of the two-state model). r_m gives the point where the binding probability reaches a maximum (crater landscape), or else values close to 1 (mesa landscape). r _s approximately indicates the onset of selection, i.e. a binding probability appreciably different from zero. (c) Adaptive dynamics as a function of time t measured in units of 1/(2s μ N) in the crater landscape at strong selection (sN = 100). Single history r(t) (dashed lines), ensemble average (thick solid lines) and width given by the standard deviation curves ± δ r(t) (thin solid lines), (d) Same as (c) in the mesa landscape at moderate (sN = 6.8) selection, (e) Stationary ensembles P_stat(r) of binding site sequences with in the crater landscape at strong selection (filled bars) and for neutral evolution (empty bars). (f) Same as (e) in the mesa landscape at moderate selection, together with the histogram of Hamming distances of CRP site sequences in E. coli from their consensus sequence (diamonds, from [10]).