Table 2. Variance in viral load (VL) explained by linear regression analysis with different models.

PC1-PC4 denote first four principal components of principal component analysis performed on consensus AA variants to account for the viral population structure. The 16 and 13 HIV AA positions are all positions significantly associated with VL when position-wise Shannon entropy or consensus variants, respectively, are used to explain VL.

Model	% of VL variance explained
VL ~ Sample’s mean Shannon entropy (all samples)	4.14
VL ~ Sample’s mean Shannon entropy (samples from <6 months infection duration)	3.91
VL ~ Sample’s mean Shannon entropy (samples from 6–24 months infection duration)	3.56
VL ~ Sample’s mean Shannon entropy (samples from >24 months infection duration)	10.7
VL ~ Average sequencing depth	0.68
VL ~ Sample’s mean Shannon entropy + age + sex + race + duration of infection + [PC1-PC4]	10.3
VL ~ Sample’s mean Shannon entropy + age + sex + race + duration of infection + [PC1-PC4] + average sequencing depth	10.7
VL ~ Shannon entropy at 16 HIV AA positions + age + sex + race + duration of infection + [PC1-PC4]	21.9
VL ~ Consensus variants at 13 HIV AA positions + age + sex + race + duration of infection + [PC1-PC4]	6.2