Table 3. Summary table comparing the Akaike information criterion (AIC) of the main model with model variants.

Model variant	AIC	Comments	Figure(s)	Table(s)
1. Basic virus dynamics model without an explicit effector response (Eqs 25–28)	4170	Model fits could not capture viral load data	S9 Fig	S4 Table
2. Model of post-ART control (Eqs 31–35)	4154	Model fits could not capture viral load data	S10 Fig	S5 Table
3. Hill coefficient, n = 3	4068	Model fits do not consistently capture reestablishment of control after effector depletion (see macaques MVJ and DEWP)	S11 Fig	S6 Table
4. Hill coefficient, n = 1	4223	Model fits could not capture viral load data	S12 Fig	S7 Table
5. Main model (Eqs 13–20)	3998		Figs 2, 3 and 4	Tables 1 and 2, S1 and S2 Tables
6. No enhanced antigen clearance by bNAbs (no AV term in Eq 15)	4152	Model fits could not capture viral load data	S13 Fig	S8 Table
7. No enhanced antigen uptake and effector elicitation by bNAbs (no f*AV term in Eq 16)	4008	Model fits do not capture the rebound viremic peak (after bNAb clearance) as well as the main model	S14 Fig	S9 Table
8. The effector activation and exhaustion thresholds are not same (ϕ1 ≠ ϕ2)	4078	Model fits did not capture viral load data better than the main model with ϕ1 = ϕ2 = ϕ	−	−
9. Varying the rate at which effectors proliferate (kE in Eq 16)	4010	Model fits did not capture viral load data better than the main model with fixed kE = 0.1 day-1	S8 Fig	S3 Table