Design and characterization of novel immunogens for AIDS vaccine development and evaluation of a sample inference method for NGS Illumina amplicon data
Since the beginning of the AIDS pandemic, an estimated 78 million people have become infected and 35 million people have died from AIDS-related illnesses. Despite the existence of effective antiretroviral therapy, 1.1 million people died of AIDS-related causes in 2015. A vaccine that could induce broadly neutralizing antibodies (bnAbs) is hypothesized to be the most efficient way to halt the AIDS pandemic. However, the majority of attempts to elicit bnAbs with HIV-1 vaccine candidates have failed due to the extensive variability and complex immune-evasion strategies of HIV-1. Recent advances in the isolation of bnAbs from HIV-1 infected individuals have revived interest in vaccine development. The membrane proximal external region (MPER) of gp41 and the CD4 binding site (CD4bs) of gp120 have become attractive targets for vaccine development because they contain highly conserved epitopes recognized by some of the broadest neutralizing antibodies. Here, we have designed and characterized multiple immunogens and vaccine strategies to induce bnAbs targeted to MEPR or CD4bs. Our findings indicate that 1) neighboring domains influence the immunogenicity of gp41 MPER, and 2) priming with a small gp41 or gp120 immunogen, then subsequently boosting with larger and more native immunogens, may have the potential to elicit antibodies towards the appropriate neutralizing epitopes.
Illumina amplicon sequencing is an important tool for the identification and quantification of species or variants in metagenomics studies, but sequencing errors make it challenging to correctly identify the authentic differences. Many denoising algorithms have been developed, but most ignore the quality scores or compress that data. We developed ampliclust, an error modeling approach using uncompressed sequences and quality scores to infer samples in Illumina amplicon data. Our approach showed better accuracy than the popular denoising tool DADA2 when data are not well separated.