.png){kind=icon}
.png){kind=icon}
Return
Nanotrap–AI Integration Enables Ultra-Sensitive Point-of-Care HIV Testing
.png){kind=icon}
PRE
AI
J
S
H
J
D
S
J
DOI：10.1021/acsnano.6c01808.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Early HIV detection using noninvasive samples remains challenging because oral fluid contains extremely low antibody levels and enzymatic inhibitors that limit the sensitivity of lateral-flow assays (LFAs). We introduce BE-SMART-HIV, a diagnostic platform that integrates a BEETLES2-inspired bioengineered enrichment (BE) nanotrap with a smartphone-based deep-learning reader (SMART). The BE nanotrap concentrates antibodies by ∼20-fold while removing salivary inhibitors, enabling oral-fluid-equivalent samples to become detectable on commercial LFAs. A transfer-learning-refined AI model interprets weak test lines with 98.6% accuracy, outperforming untrained users and clinicians. In a longitudinal seroconversion panel, BE-SMART-HIV detected antibody emergence up to 8 days earlier than conventional LFAs and reproduced ELISA-like temporal patterns, including IgM onset and IgG maturation, even from samples diluted 1,000-fold. These findings demonstrate that enriched oral-fluid-level specimens can capture systemic antibody kinetics, establishing a practical route to early, noninvasive, high-fidelity HIV screening for repeated testing in resource-limited settings.
Keywords:
deep learning
nanotrap
sample preparation
LFA
HIV