Dimensionality-reduction-based visualization is essential for interpreting complex biological data. Yet, unsupervised methods such as t-distributed stochastic neighbor embedding, Uniform Manifold Approximation and Projection, and Isomap reflect only the dominant data structure, which may not align with the goals of downstream analysis or expert-provided annotations. Existing supervised variants only partially address this mismatch and introduce new limitations. Here we present RF-PHATE, a supervised visualization approach that incorporates expert knowledge to reveal label-relevant structure while suppressing extraneous variation. RF-PHATE uses random forests to learn relationships between features and labels and translates this information into low-dimensional embeddings. RF-PHATE handles large datasets and is suitable for both classification and regression tasks. We demonstrate its use across four case studies, including longitudinal multiple sclerosis data, Raman spectral measurements of antioxidant effects, outcomes of patients with COVID-19, and RNA sequencing data with simulated dropout. These applications highlight RF-PHATE's ability to enhance interpretability, manage noise and expose meaningful biological structure, suggesting broad potential for improving data exploration and discovery.

Simian immunodeficiency virus (SIV) primarily infects CD4+ T cells, yet the relative contribution of different subsets to viral production in blood and tissues remains unclear. We develop SIV-Flow, a flow cytometry-based assay, to quantify and characterize the phenotype of productively infected cells in blood, spleen, and multiple lymph nodes from eight chronically SIV-infected rhesus macaques. We measured a consistent average of ∼1,000 infected cells per million CD4+ T cells, with limited variation by the anatomical site. Most productively infected cells displayed a memory phenotype, dominated by the effector memory subset. Despite modest tissue-specific differences, infected cells were uniformly enriched for the activation (PD-1 and CD69), integrin (VLA-4), and proliferation (Ki-67) markers. A unique subset of memory cells expressing high levels of Ki67 and PD-1 represented 44% of all infected cells. These data identify proliferating effector memory CD4+ T cells as key contributors to viral production and suggest that their phenotypic profile is conserved across diverse tissues.

Early initiation of antiretroviral therapy (ART) in perinatally HIV-infected children significantly limits the establishment of the viral reservoir. However, the long-term impact of this intervention remains unclear. We measured the frequency of inducible, translation-competent, and replication-competent proviruses in samples from 62 children who initiated ART early and remained virally suppressed for up to 9.9 years. Only a small fraction of HIV genomes produced HIV transcripts, viral proteins, or infectious virions. Accordingly, replication-competent virus was detected in only 11% of the participants. Despite the predominance of naive cells in pediatric blood, most proviruses were detected in memory CD4+ T cells, especially central memory cells. Longitudinal analysis revealed a biphasic decay in HIV DNA: an initial decline followed by long-term stability, which was associated with extensive expansions of infected T cell clones. In contrast, inducible proviruses declined continuously and became undetectable in most children after 5 years. Near full-length sequencing of 1,305 HIV genomes revealed a dramatic reduction in genetically intact proviruses, from pre-ART to after 7 years of ART. Together, these findings suggest that the intact viral reservoir rapidly decays in early-treated children, offering critical insights for pediatric HIV cure strategies.

Background. People with human immunodeficiency virus (HIV) receiving antiretroviral therapy (ART) have increased risks of non-AIDS comorbidities. Growth differentiation factor 15 (GDF-15) is a mitokine released upon mitochondrial stress and is a validated aging biomarker. Herein, we assessed associations between plasma GDF-15 levels, inflammation, and HIV reservoir markers in ART-treated people with HIV (PWH). Methods. Blood samples were collected from 78 ART-naive, 140 ART-treated PWH (median ART duration, 15.6 years) and 83 uninfected control participants. GDF-15 and markers of inflammation were quantified in plasma by enzyme-linked immunosorbent assay (ELISA) and multiplex assays. Integrated HIV DNA levels were measured in isolated CD4 T cells by ultrasensitive quantitative Alu polymerase chain reaction. Intracellular GDF-15 production was assessed ex vivo by flow cytometry, and in supernatants by ELISA after in vitro stimulations. Results. Plasma GDF-15 levels were higher in ART-treated PWH compared to ART-naive PWH or controls, independently of age. Ex vivo, GDF-15 was produced by classical, intermediate, and nonclassical monocytes, but absent in T cells, B cells, natural killer cells, and dendritic cells. Plasma and monocyte intracellular GDF-15 levels correlated strongly (r = 0.96, P = .002). Plasma GDF-15 levels did not correlate with the majority of inflammatory markers quantified in plasma. Conversely, plasma GDF-15 levels correlated with the validated non-AIDS inflammatory comorbidity marker suPAR (soluble urokinase plasminogen activator receptor; r = 0.59, P < .001), as well as with integrated HIV DNA levels in CD4+ T cells (r = 0.47, P < .01). Conclusions. Plasma levels of GDF-15, mainly produced by monocytes, were positively associated with HIV reservoir markers and coincided with increased level of suPAR, suggesting that HIV persistence is associated with increased mitochondrial stress and comorbidity risks in ART-treated PWH.

The SARS-CoV-2 Spike (S) protein is essential for viral entry and serves as the primary immunogen in most COVID-19 vaccines. While its role in adaptive immunity is well defined, its potential to contribute directly to innate immune activation remains incompletely understood. Neutrophils, in particular, are prominent effectors in COVID-19 severity, yet how they respond directly to the S protein presented in a multivalent format is unclear. Here, we investigated whether the S protein can directly activate human neutrophils ex vivo using two biologically relevant models: nanoparticles displaying multivalent stabilized prefusion trimeric S glycoprotein, and purified β-propiolactone-inactivated SARS-CoV-2 virions. Neutrophils were exposed to nanoparticles or inactivated virus, either alone or pre-coated with monoclonal or polyclonal anti-S antibodies. Nanoparticles displaying Respiratory Syncytial Virus (RSV) Fusion (F) protein and purified β-propiolactone-inactivated RSV served as comparators. Across all models and conditions tested, the S protein did not induce significant neutrophil responses. No consistent effects were observed on cell viability, surface marker expression, reactive oxygen species production, neutrophil extracellular trap formation, cytokine release, or inflammatory gene expression-even in the presence of anti-S antibodies mimicking immune complexes. Results with F-nanoparticles and inactivated RSV were similarly modest. These findings indicate that the trimeric prefusion S protein, whether displayed multivalently on nanoparticles or in the context of inactivated viral particles, is insufficient to trigger robust neutrophil activation. This work provides insight into the innate immune profile of the S protein and suggests that its use in vaccine platforms is unlikely to directly provoke neutrophil-mediated inflammatory responses.

Persistent SIV/HIV reservoirs are the primary obstacle to a cure and the source of viral rebound after ART interruption (ATI). However, the anatomical source of viral rebound remains elusive. Here, we characterized the proviral landscape in the blood, inguinal, and axillary lymph nodes and colon biopsies of five SHIV-infected rhesus macaques (RMs), under ART for 28 weeks. From the 144 near full-length (NFL) proviral sequences obtained pre-ATI, 35% were genetically intact and only 2.8% were found in multiple copies. Envelope sequences of plasma rebounding viruses after ATI, more frequently matched pre-ATI intact proviruses retrieved from lymph nodes compared to sequences isolated from the blood or the colon (4, 1, and 1 pair of matched sequences, respectively). Our results suggest that clonal expansion of infected cells rare in this model, and that intact proviruses persisting in the lymph nodes may be a preferential source of viral rebound upon ATI.

HIV persistence despite years of ART suppression poses a major barrier to cure. Using a full-length latency reporter to generate HIV-infected, transcriptionally silent CD4+ T cells in vitro , we show that cognate DC:T cell interactions drive clonal expansion of latent T cells in an antigen dependent manner and that a pro-survival state within proliferating cells is reinforced through IL-7 signaling. Interestingly, we describe a dominant role for CD28 co-stimulation in regulating robust latent T cell proliferation which was partially reversed by PD-1 blockade. Our studies show that a gradual reduction in antigenic stimulation was sufficient to induce proliferative responses without measurable proviral reactivation. Thus, the magnitude of TCR/co-stimulatory signals during cognate APC:T cell interactions are key regulators of the underlying proliferative and survival programs maintaining the latent reservoir under ART suppression. Significance Statement Viral replication can be effectively suppressed by antiretroviral therapy (ART) but is not curative due to persistence of latent virus in a stable reservoir in resting CD4+ T cells. We show that antigen recognition through cell-cell interactions is an important driver of latent T cell proliferation, and that modulating TCR stimulatory signaling independently regulates proliferative, survival and proviral reactivation potential in infected T cells. Our observations show that latent T cells retain their ability to engage other immune cells to support their long-term survival under ART suppression, similar to uninfected T cells. These characteristics of latent T cell pools represent an additional hurdle to eradicating the reservoir. ### Competing Interest Statement The authors have declared no competing interest. Canadian Institutes of Health Research, https://ror.org/01gavpb45, 155951, HIG-133050

ABSTRACT Anti-HIV-1 antibodies capable of mediating ADCC are elicited by the majority of people with HIV-1 and preferentially target the “open,” CD4-bound conformation of HIV-1 envelope glycoproteins (Env). However, due to the “closed” conformation sampled by unliganded HIV-1-Envs, these antibodies are ineffective at eliminating infected cells. BNM-III-170 is a small-molecule CD4-mimetic compound that binds the Phe43 cavity of the gp120 subunit of Env, forcing Env to “open up,” thus exposing epitopes targeted by CD4-induced (CD4i), ADCC-mediating antibodies. Here, we assessed the safety, pharmacokinetics, and biological activity of BNM-III-170 in uninfected and SHIV-AD8-EO-infected rhesus macaques (RMs). In uninfected RMs, single subcutaneous administrations of 3–36 mg/kg BNM-III-170 were well-tolerated, with serum half-lives ranging from 3 to 6 h. In SHIV-infected RMs, four different regimens were evaluated: 2 × 36 mg/kg daily, 1 × 24 mg/kg, 3 × 36 mg/kg every 7 days, and 3 × 36 mg/kg every 3 days. While toxicity was observed with daily doses, all other regimens demonstrated reasonable safety profiles. No changes in plasma viral loads were observed in SHIV-infected RMs following any of the evaluated BNM-III-170 dosing regimens. However, plasma collected following BNM-III-170 administration was shown to have increased binding to infected cells and to sensitize SHIV AD8-EO virions to neutralization by otherwise non-neutralizing antibodies. In addition, the plasma of treated animals mediated ADCC in the presence of BNM-III-170. These results establish a well-tolerated BNM-III-170 dosing regimen in SHIV-infected RMs and serve as proof of concept for its biological activity in promoting the targeting of infected cells by CD4i ADCC-mediating antibodies. Thus, they inform future studies evaluating CD4mc treatment in ART-treated animals. IMPORTANCE A therapeutic regimen able to eradicate or functionally cure HIV-1 remains elusive and may require a “shock-and-kill” approach to reactivate and then purge the latent HIV-1 reservoir. The small-molecule CD4-mimetic compound BNM-III-170 has previously been shown to (i) sensitize HIV-1-infected cells to ADCC mediated by plasma from people with HIV-1 (PWH) in vitro and (ii) significantly delay the time to viral rebound following ART interruption when combined with anti-CoRBS + anti-cluster A Abs or plasma from PWH in humanized mice. To evaluate the use of BNM-III-170 as part of a kill approach, we characterized the safety, pharmacokinetics, and biological activity of BNM-III-170 in uninfected and SHIV-infected RMs. Our study identifies a tolerable BNM-III-170 dosing regimen in SHIV-infected RMs and provides insights into its antiviral activities; as such, it informs future studies evaluating the efficacy of BNM-III-170 in reducing the viral reservoir.

Interleukin 32 (IL-32) is a potent multi-isoform proinflammatory cytokine, which is upregulated in people with HIV (PWH) and is associated with cardiovascular disease (CVD) risk. However, the impact of IL-32 isoforms on CD4 T-cell cardiotropism, a mechanism potentially contributing to heart inflammation, remains unknown. Here we show that IL-32 isoforms β and γ induce the generation of CCR4+CXCR3+ double positive (DP) memory CD4 T-cell subpopulation expressing the tyrosine kinase receptor c-Met, a phenotype associated with heart-homing of T cells. Our ex vivo studies on PWH show that the frequency of DP CD4 T cells is significantly higher in individuals with, compared to individuals without, subclinical atherosclerosis and that DP cells from antiretroviral-naive and treated individuals are highly enriched with HIV DNA. Together, these data demonstrate that IL-32 isoforms have the potential to induce heart-homing of HIV-infected CD4 T cells, which may further aggravate heart inflammation and CVD in PWH.

Background Reemergence of human herpesvirus 8 (HHV-8)-induced Kaposi sarcoma (KS) in people living with HIV (PLWH) despite antiretroviral therapy (ART) poses a clinical challenge because they already have favorable CD4 T-cell numbers and undetectable viral loads. We observed that clinical presentation in PLWH on ART resembled classic KS found in older HIV-uninfected patients and hypothesized that immunosenescence may thus play a role in occurrence of KS on ART. We compared viral and immune factors implicated in the development of KS in ART-treated PLWH (HIV KS) and HIV-uninfected classic KS patients (cKS), compared to controls without KS (HIV Control, cControls respectively).Methods Plasma, peripheral blood mononuclear cell, and skin tissues were obtained from 11 HIV KS and 11 cKS patients and 2 groups of age-matched controls.Results HIV KS participants were younger than cKS (aged 53 vs 75 years). HHV-8 genotypes did not differ between groups. Despite the younger age and a lower CD4/CD8 ratio, activated, exhausted, and senescent T-cell frequencies were similar between HIV KS and cKS. Anti-HHV-8 immunoglobulin G levels were higher and circulating HHV-8 DNA lower in HIV KS compared with cKS. Circulating platelet-derived growth factors AA-BB and granulocyte colony-stimulating factors were higher in HIV KS We observed similar levels of HHV-8 DNA and PD-1 expression in skin lesions from HIV KS and cKS patients.Conclusions Altogether, early immune senescence could be involved in the development of KS in ART-treated PLWH. Higher anti-HHV-8 immunoglobulin G levels could be linked with lower circulating viral load. Such insights should help developing therapeutical strategies to prevent development and treat KS in PLWH on ART.

Throughout the SARS-CoV-2 pandemic, several variants of concern (VOCs) have been identified, many of which share recurrent mutations in the spike glycoprotein’s receptor-binding domain (RBD). This region coincides with known epitopes and can therefore have an impact on immune escape. Protracted infections in immunosuppressed patients have been hypothesized to lead to an enrichment of such mutations and therefore drive evolution towards VOCs. Here, we present the case of an immunosuppressed patient that developed distinct populations with immune escape mutations throughout the course of their infection. Notably, by investigating the co-occurrence of substitutions on individual sequencing reads in the RBD, we found quasispecies harboring mutations that confer resistance to known monoclonal antibodies (mAbs) such as S:E484K and S:E484A. These mutations were acquired without the patient being treated with mAbs nor convalescent sera and without them developing a detectable immune response to the virus. We also provide additional evidence for a viral reservoir based on intra-host phylogenetics, which led to a viral substrain that evolved elsewhere in the patient’s body, colonizing their upper respiratory tract (URT). The presence of SARS-CoV-2 viral reservoirs can shed light on protracted infections interspersed with periods where the virus is undetectable, and potential explanations for long-COVID cases.

Plasma RNAemia, delayed antibody responses and inflammation predict COVID-19 outcomes, but the mechanisms underlying these immunovirological patterns are poorly understood. We profile 782 longitudinal plasma samples from 318 hospitalized patients with COVID-19. Integrated analysis using k-means reveals four patient clusters in a discovery cohort: mechanically ventilated critically-ill cases are subdivided into good prognosis and high-fatality clusters (reproduced in a validation cohort), while non-critical survivors segregate into high and low early antibody responders. Only the high-fatality cluster is enriched for transcriptomic signatures associated with COVID-19 severity, and each cluster has distinct RBD-specific antibody elicitation kinetics. Both critical and non-critical clusters with delayed antibody responses exhibit sustained IFN signatures, which negatively correlate with contemporaneous RBD-specific IgG levels and absolute SARS-CoV-2-specific B and CD4+ T cell frequencies. These data suggest that the "Interferon paradox" previously described in murine LCMV models is operative in COVID-19, with excessive IFN signaling delaying development of adaptive virus-specific immunity. The role of IFN signaling in SARS-CoV-2 infection and outcome is still debated. Here, the authors longitudinally profiled plasma samples from hospitalized patients and show that a persistent inflammatory response is linked to delayed generation of adaptive immunity and increased risk of death when coupled with severe infection.

Background Despite successful antiretroviral therapy (ART), frequencies and immunological functions of memory CCR6+ Th17-polarised CD4+ T-cells are not fully restored in people with HIV (PWH). Moreover, long-lived Th17 cells contribute to HIV persistence under ART. However, the molecular mechanisms underlying these observations remain understudied. Methods mRNA-sequencing was performed using Illumina technology on freshly FACS-sorted memory CCR6+CD4+ T-cells from successfully ART-treated (ST), elite controllers (EC), and uninfected donors (HD). Gene expression validation was performed by RT-PCR, flow cytometry, and in vitro functional assays. Findings Decreased Th17 cell frequencies in STs and ECs versus HDs coincided with reduced Th17-lineage cytokine production in vitro. Accordingly, the RORγt/RORC2 repressor NR1D1 was upregulated, while the RORγt/RORC2 inducer Semaphorin 4D was decreased in memory CCR6+ T-cells of STs and ECs versus HDs. The presence of HIV-DNA in memory CCR6+ T-cells of ST and EC corresponded with the downregulation of HIV restriction factors (SERINC3, KLF3, and RNF125) and HIV inhibitors (tetraspanins), along with increased expression of the HIV-dependency factor MRE11, indicative of higher susceptibility/permissiveness to HIV-1 infection. Furthermore, markers of DNA damage/modification were elevated in memory CCR6+ T-cells of STs and ECs versus HDs, in line with their increased activation (CD38/HLA-DR), senescence/exhaustion phenotype (CTLA-4/PD-1/CD57) and their decreased expression of proliferation marker Ki-67. Interpretation These results reveal new molecular mechanisms of Th17 cell deficit in ST and EC PWH despite a successful control of HIV-1 replication. This knowledge points to potential therapeutic interventions to limit HIV-1 infection and restore frequencies, effector functions, and senescence/exhaustion in Th17 cells. Funding This study was funded by the Canadian Institutes of Health Research (CIHR, operating grant MOP 142294, and the Canadian HIV Cure Enterprise [CanCURE 2.0] Team Grant HB2 164064), and in part, by the Réseau SIDA et maladies infectieuses du Fonds de recherche du Québec-Santé (FRQ-S).

Dimensionality reduction-based data visualization is pivotal in comprehending complex biological data. The most common methods, such as PHATE, t-SNE, and UMAP, are unsupervised and therefore reflect the dominant structure in the data, which may be independent of expert-provided labels. Here we introduce a supervised data visualization method called RF-PHATE, which integrates expert knowledge for further exploration of the data. RF-PHATE leverages random forests to capture intricate featurelabel relationships. Extracting information from the forest, RF-PHATE generates low-dimensional visualizations that highlight relevant data relationships while disregarding extraneous features. This approach scales to large datasets and applies to classification and regression. We illustrate RF-PHATE’s prowess through three case studies. In a multiple sclerosis study using longitudinal clinical and imaging data, RF-PHATE unveils a sub-group of patients with non-benign relapsingremitting Multiple Sclerosis, demonstrating its aptitude for time-series data. In the context of Raman spectral data, RF-PHATE effectively showcases the impact of antioxidants on diesel exhaust-exposed lung cells, highlighting its proficiency in noisy environments. Furthermore, RF-PHATE aligns established geometric structures with COVID-19 patient outcomes, enriching interpretability in a hierarchical manner. RF-PHATE bridges expert insights and visualizations, promising knowledge generation. Its adaptability, scalability, and noise tolerance underscore its potential for widespread adoption.

This study provides new insights into the contributions of different CD4 + and CD8 + T-cell subsets to the anatomic differences between LN and blood in individuals with HIV who have optimal versus suboptimal CD4 + T-cell recovery. To our knowledge, this is the first study comparing paired LN and blood CD4 + and CD8 + T-cell differentiation subsets, as well as those subsets in immunological responders versus immunological suboptimal responders.

ABSTRACT The human immunodeficiency virus (HIV) persists in HIV viral reservoirs despite antiretroviral therapy (ART). Hepatitis C virus (HCV) coinfection is associated with increased HIV reservoir size and residual HIV transcription during ART. Herein, we investigated the impact of direct acting antivirals (DAA)-mediated HCV cure on the size/transcriptional activity of the HIV reservoirs and investigated predictors of HIV reservoirs decline in HCV+/HIV+ coinfected individuals. HCV+/HIV+ ( n = 20) and HCV+/HIV− ( n = 14) participants were examined prior to DAA treatment (baseline), at the end of treatment (EOT), and at 12–24 weeks after EOT (follow-up). In HCV+/HIV+ individuals, DAA-mediated HCV cure significantly reduced integrated HIV DNA levels, mainly in participants infected with HCV prior to HIV. Integrated HIV DNA, unspliced (US), and multiply spliced (MS) HIV RNA levels were quantified in sorted CD4+ T-cells. Despite the transient elevation of US HIV RNA at EOT, changes in US and MS HIV RNA were not statistically significant. Plasma inflammation markers were measured and DAA also reduced plasma sCD163 and sCD14. Changes in immunological/virological parameters were analyzed using multivariate random forest analysis where pre-ART peak HIV viremia and HCV viral load predicted integrated HIV DNA and US RNA changes. In conclusion, we demonstrate the beneficial impact of DAA on HIV reservoirs and immune activation, with a fraction of HIV reservoirs being DAA-sensitive in people infected with HCV before HIV. Furthermore, we identify HIV/HCV viremia as the top predictors of DAA-mediated changes in the HIV reservoirs. These findings support the need for early ART and DAA treatment in HIV/HCV coinfections. IMPORTANCE Antiretroviral therapy (ART) for human immunodeficiency virus (HIV) can control virus replication and prolong the life of people living with HIV (PLWH). However, the virus remains dormant within immune cells in what is called the HIV reservoir. Furthermore, 2.3 million PLWH are also coinfected with hepatitis C virus (HCV) and are at risk of developing chronic liver disease and cancer. HCV treatment with direct acting antivirals (DAA) can completely cure the infection in more than 95% of treated individuals and improve their long-term health outcomes. In this study, we investigated how HCV treatment and cure affect the HIV reservoir. We demonstrate the beneficial impact of DAA treatment as it reduces the HIV reservoirs in particular in people infected with HCV before HIV. These results support the need for early ART and DAA treatment in HIV/HCV coinfections.

Spontaneous transcription and translation of HIV can persist during suppressive antiretroviral therapy (ART). The quantity, phenotype, and biological relevance of this spontaneously "active"reservoir remain unclear. Using multiplexed single-cell RNAflow-fluorescence in situ hybridization (FISH), we detect active HIV transcription in 14/18 people with HIV on suppressive ART, with a median of 28/million CD4(+) T cells. While these cells predominantly exhibit abortive transcription, p24(-)expressing cells are evident in 39% of participants. Phenotypically diverse, active reservoirs are enriched in central memory T cells and CCR6(-)and activation marker-expressing cells. The magnitude of the active reservoir positively correlates with total HIV-specific CD4(+) and CD8(+) T cell responses and with multiple HIV-specific T cell clusters identified by unsupervised analysis. These associations are particularly strong with p24-expressing active reservoir cells. Single-cell vDNA sequencing shows that active reservoirs are largely dominated by defective proviruses. Our data suggest that these reservoirs maintain HIV-specific CD4(+) and CD8(+)T responses during suppressive ART.

Before initiation of antiretroviral therapy (ART), HIV-specific CD8+T cells are dysfunctional and short lived. To better understand the relationship between the HIV reservoir in CD4+ T cells and the magnitude and differentiation status of HIV-specific CD8+ T cells, we investigated these cells from acute and chronic HIV-infected individuals after 2 years of ART. Although both the HIV reservoir and the CD8+ T cell responses declined significantly after 2 years of ART, sustained HIV-specific CD8+ T cell responses correlated with a greater reduction of integrated HIV provirus. However, the magnitude of CD8+ T cells specific for HIV Gag, Pol, Nef, and Vif proteins positively associated with the active reservoir size during ART, measured as cell -associated RNA. Importantly, high HIV DNA levels strongly associate with maintenance of short-lived HIV-specific CD8+ T cells, regardless of ART initiation time. Our data suggest that the active reservoir maintains HIV -specific CD8+ T cell magnitude but prevents their differentiation into functional cells.

Objectives. Identifying biomarkers causing differential SARS-CoV-2 infection kinetics associated with severe COVID-19 is fundamental for effective diagnostics and therapeutic planning. Methods. In this work, we applied mathematical modelling to investigate the relationships between patient characteristics, plasma SARS-CoV-2 RNA dynamics and COVID-19 severity. Using a straightforward mathematical model of within-host viral kinetics, we estimated key model parameters from serial plasma viral RNA (vRNA) samples from 256 hospitalised COVID-19(+) patients. Results. Our model predicted that clearance rates distinguish key differences in plasma vRNA kinetics and severe COVID-19. Moreover, our analyses revealed a strong correlation between plasma vRNA kinetics and plasma receptor for advanced glycation end products (RAGE) concentrations (a plasma biomarker of lung damage), collected in parallel to plasma vRNA from patients in our cohort, suggesting that RAGE can substitute for viral plasma shedding dynamics to prospectively classify seriously ill patients. Conclusion. Overall, our study identifies factors of COVID-19 severity, supports interventions to accelerate viral clearance and underlines the importance of mathematical modelling to better understand COVID-19.