Dual-Layer Serological Encoding: Integrating Pathogen-Derived and Host-Reactive Peptide Signatures Reveals Complementary Dimensions of HIV Immune Response

Abstract

Serological diagnostics traditionally rely on pathogen-derived antigens to detect infection-specific antibody responses. Chronic infections also induce systemic immune remodeling that may be reflected in global antibody reactivity patterns beyond antigen specificity. Here we evaluate a dual-layer serological framework combining HIV-derived peptides with a host-derived peptide library designed to capture distributed antibody reactivity patterns. Using strict nested cross-validation in a cohort of 105 individuals, pathogen-derived 12-mer peptides achieved high classification performance with an AUC of 0.891, whereas the 10-mer host-based peptide library alone yielded moderate but statistically significant discrimination with an AUC of 0.805. Integration via regularized stacking resulted in only a modest additive improvement, reaching an AUC of 0.897, indicating partial redundancy in diagnostic ranking. In contrast, entropy and inequality analyses revealed substantial immune repertoire restructuring in HIV-positive individuals, characterized by reduced Shannon entropy and significant correlations between classifier probability and repertoire concentration. These findings support a dual-layer model of serology in which the integration of pathogen-derived and host-derived peptides into a meta model encode antigen specificity, whereas host-reactive signatures reflect systemic immune topology. Distinguishing diagnostic ranking from immune-state encoding provides a conceptual framework for multi-layer serological diagnostics.