Abstract
We formalize three models of genome evolution, yielding analytical distributions of synteny block sizes between pairs of genomes. To evaluate each model, we fit theoretical synteny block size distributions to empirical distributions from a comparative analysis of nearly 200 mammalian genomes. Between the random breakage model (RBM), with uniformly distributed breakpoints, the fragile breakage model (FBM) with breakpoint hotspots, and what we term the correlated breakage model (CBM) with pairs of nearby breakpoints, we find that the FBM and CBM are far more consistent with mammalian genome evolution than the RBM. Inferred parameters of each model qualitatively agree with previous characterizations of mammalian genome evolution. The FBM and CBM perform comparably with each other, suggesting that either model could serve as an improved null model of genome evolution to use in statistical tests for syntenic regions that are unlikely to exist due to random chance.