Table 2. Cox proportional hazards models quantify contributions to escape rate.

Predictors	Coefficient	P value	Pseudo-R2
Univariate models (n=53 epitopes, maximum possible pseudo-R2=0.99)
log10(S)	0.87	0.08	0.06
ΔE	−0.14	0.02	0.10
tWF	−0.14	5.8 × 10−8	0.51
tWF%M	−0.17	1.5 × 10−9	0.63
log10(%M)	1.53	7.8 × 10−5	0.29
Multivariate models (n=53, maximum possible pseudo-R2=0.99)
log10(S)+ log10(%M)	1.111.60	0.079.3 × 10−5	0.33
ΔE+ log10(%M)	−0.171.66	4.4 × 10−33.6 × 10−5	0.39
tWF+ log10(%M)	−0.141.55	1.3 × 10−71.7 × 10−4	0.64
tWF%M+ log10(%M)	−0.160.13	1.7 × 10−70.76	0.64
Univariate models, excluding escapes at the time the T-cell response was first detected (n=49, maximum possible pseudo-R2=0.99)
log10(S)	0.81	0.11	0.05
ΔE	−0.14	0.02	0.10
tWF	−0.12	8.9 × 10−6	0.37
tWF%M	−0.15	1.1 × 10−7	0.53
log10(%M)	1.68	5.1 × 10−5	0.33
Multivariate models, excluding escapes at the time the T-cell response was first detected (n=49, maximum possible pseudo-R2=0.99)
log10(S)+ log10(%M)	1.061.77	0.106.2 × 10−5	0.37
ΔE+ log10(%M)	−0.181.83	5.0 × 10−32.3 × 10−5	0.42
tWF+ log10(%M)	−0.121.65	2.0 × 10−51.1 × 10−4	0.56
tWF%M+ log10(%M)	−0.140.37	4.8 × 10−50.43	0.54

Contributions of vertical immunodominance (%M) and purely fitness-related measures (S, ΔE and t_WF) are mostly independent.