Table 1.

Overview of model fits

				fitting error					fluidity
	G	M	ΔA	ΔB	ΔC	ΔD	φobs	${\phi }_{\mathsf{\text{A}}}^{\mathsf{\text{pred}}}$	${\phi }_{\mathsf{\text{B}}}^{\mathsf{\text{pred}}}$	${\phi }_{\mathsf{\text{C}}}^{\mathsf{\text{pred}}}$	${\phi }_{\mathsf{\text{D}}}^{\mathsf{\text{pred}}}$
B. anthracis	13	5523	80	21	78	13	0.08	0.09	0.08	0.09	0.08
E. coli	15	4576	98	58	47	2.6	0.25	0.30	0.25	0.29	0.25
Staph. aureus	19	2651	29	16	21	4.3	0.16	0.19	0.16	0.19	0.16
Strep. pneumonia	26	2095	42	21	30	4.3	0.23	0.32	0.24	0.30	0.23
Strep. pyogenes	14	1786	26	10	25	7.5	0.20	0.24	0.20	0.24	0.21
N. meningitidis	12	2080	53	26	31	2.4	0.28	0.33	0.28	0.32	0.28

Model A assumes a constant population size, and the same gene transfer process for all genes. Model B assumes an exponentially growing population size. Model C assumes that a part of the genome is shared by all genomes (a rigid core); the other part is subjected to the same gene transfer process as in model A. Model D assumes two parts in the genomes, governed by different gene transfer rates. We determined for the four models the parameters that minimize the distance Δ between the empirical and the theoretical gene frequency distribution (see Materials and Methods for the definition of Δ). For each of the 6 bacterial species analyzed, we report the number of analyzed genomes G, the genome size M (average number of genes per genome), the distance Δ for the model fits, the genomic fluidity φ^obs estimated on the data, and the fluidity φ^pred for the model fits. Recall that model A has one parameter, models B and C have two parameters, and model D has three parameters.