Figure 4. A selective sweep of spacer diversity.

(A) Hosts CRISPR loci from the simulation in Figure 3 are clustered (Text S3) into distinct sub-populations every 100 model iterations to capture how trailer-end clonality emerges. Cluster heights represent the cumulative frequencies of all strains in a given cluster, cluster widths show the number of distinct strains in that cluster, and the combined height of all clusters in an iteration reflects the fraction of virus-host interactions that is immune (i.e., host mean fitness). A marked loss of host diversity occurs prior to iteration 3800 (→), after which the sweeping sub-population diversifies through distinct leader-end spacer incorporations (Figure S6). (B) The frequencies of all host spacers at a single leader-end column are tracked during the clustering-predicted sweep. A single spacer (shown in black) rapidly rises in frequency before iteration 3800 as predicted by the clustering. Subsequent ‘kill the winner’ oscillations occur before all competing hosts go extinct. A second sweep purges the remaining diversity at this locus position.