Abstract

Microbial translocation is a process wherein bacteria migrate from the gut to the blood, due to an alteration of the gut barrier [1]. This phenomenon is significant as it may trigger a persistent activation of the immune system, potentially influencing vaccine or treatment response [2]. Our understanding of the impact of this process is limited necessitating the development and the use of sophisticated bioinformatics and statistical methods to analyze and quantify non-human molecules from blood or plasma samples.

Translocation can be addressed using metagenomics or metatranscriptomics strategies based on state-of-the-art-methods that are commonly applied to data specifically produced in this context: targeting bacteria and other non-human sequences during the experiment. However, a key challenge with translocation involves the re-use of clinical study data, such as from vaccination (e.g., EBOVAC2 [3]) to characterize the non-human metagenome or metatranscriptome from blood sequencing. The main issue here is to deal with millions of short sequences with a majority coming from the human genome.

Most clinical studies focus on the analysis of human data, while minimizing contamination by non-human molecules. When analyzing this data from a translocation perspective, non-human contaminants become the focus of the analysis as their composition has to be characterized. Therefore, such data requires initial filtration to remove human sequences, before analyzing the remaining small fraction (less than 1%) of non-human sequence.

A first work has been proposed by Douek et al. [2] to analyze the link between microbial composition and HIV infection treatment. In [2], the authors, exploit non-human reads within a pipeline that assembles such reads into contigs: longer sequences. Building on the work of Douek et al. and given the small quantities of non-human reads in vaccine cohorts, in this study we propose a pipeline that should allow enhancing sensitivity and specificity by directly analyzing non-human reads instead of contigs.

For this purpose, we implement a comparative approach to evaluate the most recent tools. This analysis raises some additional issues as the sequence databases to be used and the taxonomic assignment levels. Our findings will provide insights on the consistency of predictions across methods, regarding prevalent taxa in blood samples, and general patterns across patients.

These results will lay the groundwork for developing statistical models to relate bacterial compositions with vaccine responses. We aim to use models that incorporate robust methods to deal with the inherent sparsity and noise in such datasets. Our goal is to uncover how specific taxa in the blood correlate with the immune responses to vaccination, potentially explaining variations in vaccine efficacy among patients.

1. Sandler NG, Douek DC. Microbial translocation in HIV infection: causes, consequences and treatment opportunities. Nat. Rev. Microbiol. 2012;10:655‑66.
2. Nganou-Makamdop K, Talla A, Sharma AA, Darko S, Ransier A, Laboune F, et al. Translocated microbiome composition determines immunological outcome in treated HIV infection. Cell 2021;184:3899-3914.e16.
3. Blengio F, Hocini H, Richert L, Lefebvre C, Durand M, Hejblum B, et al. Identification of early gene expression profiles associated with long-lasting antibody responses to the Ebola vaccine Ad26.ZEBOV/MVA-BN-Filo. Cell Rep. 2023;42:113101.