Abstract Body

Background: Annual HIV incidence is the primary assessor for monitoring the epidemic’s rise and decline. In pursuit of an accurate assay, we have proposed a genomic incidence assay utilizing fingerprints harbored in the HIV sequence population, which showed over 95% accuracy among 182 incident and 43 chronic samples. Still, two major hurdles must be overcome before routine use in cross-sectional settings: 1) streamlining the cost and workflow, 2) ensuring proper classification between multiple-founder recent infections and chronic infections.

Methods: For enhancing the cost efficiency of the sequence-based assay, we developed a high-throughput next-generation sequencing platform; a signal-masking bioinformatics pipeline was devised to analyze 18,434 envelope gene segments (HXB2 7212-7601) obtained from 12 incident and 24 chronic patients. To give the assay power to appropriately discriminate multiple-founder recent infections from chronic infections, we formulated a mathematical model which posits the intersequence nucleotide base difference distribution of each subject’s sequence sample as a function of infection duration and the number of founder sequences. This model was tested by analyzing HIV subtype B and C samples from 40 incident subjects with multiple founder viruses.

Results: First, the cost-effective pyrosequencing platform correctly classified all 12 incident subjects (100% sensitivity) and 23 out of 24 chronic subjects (96% specificity). Our signal-masking bioinformatics pipeline yielded a process error rate of 5.8×10-4 per base. Sampling simulations showed that the biomarkers were tolerant of the two factors most likely to affect the accuracy: sequencing errors and template resampling. Second, a quantitative guideline for segregating viral lineages was provided by our mathematical model, enabling us to assess when each subject was infected. The infection periods obtained from our model estimates and from Fiebig laboratory staging showed a statistically significant linear relationship (p<0.0005), correctly identifying all 40 individuals with incident infections. Conclusions: The high-throughput platform permits the assay to be cost-effective, and when it is combined with our mathematical model, we can obtain recency signatures from the complex gene pool that arises from multiple founder viruses. Our sequence-based approach marks significant progress towards accurate determination of HIV incidence from genomic readouts measured from cross-sectional samples from a single blood draw.