Only slight impact of predicted replicative capacity for therapy response prediction

Area of Science:

• Virology
• Immunology
• Computational Biology

Background:

• Replication capacity (RC) of specific HIV isolates is implicated in unexpected treatment responses.
• The precise role of viral RC in antiretroviral therapy outcomes remains unclear.

Purpose of the Study:

• To develop and validate a method for predicting HIV replication capacity (RC) from genotype.
• To assess the impact of predicted viral RC (pRC) on viral load (VL) and CD4(+) T-cell count changes during antiretroviral therapy.

Main Methods:

• Support vector machines (SVMs) were used to train models predicting RC from approximately 300 genotype-RC pairs.
• Linear regression models analyzed the effect of pRC, treatment activity scores (TAS), and drug combinations on VL and CD4 changes in ~3,000 treatment change episodes.

Main Results:

• SVM models achieved a moderate correlation (Spearman's rho = 0.54) between measured RC and pRC.
• pRC inclusion significantly improved predictions of virological response, particularly for VL change at 180 days (rho = 0.45).
• pRC was inversely correlated with drug resistance at treatment initiation (rho = -0.38); CD4 count prediction did not significantly improve.

Conclusions:

• HIV viral RC can be predicted from genotype with moderate accuracy.
• Predicted RC offers a slight improvement in predicting virological treatment response, potentially due to its correlation with drug resistance.