Table 1. Codons under positive selection in the PML sample.

Mutations			Full length sequence set (n = 28)	Partial sequence set codons 43–287 (n = 42)
			P-value for the positive selection test
			2.5×10−7	3.5×10−6
Position	WT	Mutant	Bayes Empirical Bayes posterior probability
55	L	F	0.82	0.94
60	K	M,E,N	<0.5	0.94
265	N	D,T	<0.5	0.85
267	S	F,L	0.80	0.92
269	S	F,Y,C	1.00	1.00

VP1 sequences isolated from PML patients and random subsets of sequences isolated from healthy subjects were further analyzed using PAML [23]. We examined multiple models of sequence evolution incorporated in PAML including purely neutral model (M0), nearly neutral model (M1), model with positive selection (M2) and additional more complex models (M3–M8). We used likelihood ratio test (LRT) to compare the difference between models M1 and M2 to test for positive selection. P-values for positive selection in three datasets are shown together with Bayesian posterior probabilities for each codon position. Residues with Bayes Empirical Bayes posterior probabilities exceeding 0.5 are shown.