Biology subjects

Muema, D.

Publications and source records attributed to Muema, D..

3 recordsLinked to original sources

bioRxiv2025

Data-optimal scaling of paired antibody language models

Scaling laws for large language models in natural language domains are typically derived under the assumption that performance is primarily compute-constrained. In contrast, antibody language models (AbLMs) trained on paired sequences are primarily data-limited, thus requiring different considerations. To explore how model size and data scale affect AbLM performance, we trained 15 AbLMs across all pairwise combinations of five model sizes and three training data sizes. From these experiments, we derive an AbLM-specific scaling law and estimate that training a data-optimal AbLM equivalent of the highly performant 650M-parameter ESM-2 protein language model would require [~]5.5 million paired antibody sequences. Evaluation on multiple downstream classification tasks revealed that significant performance gains emerged only with sufficiently large model size, suggesting that in data-limited domains, improved performance depends jointly on both model scale and data volume.

immunology↗

bioRxiv2022

HIV skews the SARS-CoV-2 B cell response toward an extrafollicular maturation pathway

BackgroundHIV infection dysregulates the B cell compartment, affecting memory B cell formation and the antibody response to infection and vaccination. Understanding the B cell response to SARS-CoV-2 in people living with HIV (PLWH) may explain the increased morbidity, reduced vaccine efficacy, reduced clearance, and intra-host evolution of SARS-CoV-2 observed in some HIV-1 coinfections. MethodsWe compared B cell responses to COVID-19 in PLWH and HIV negative (HIV-ve) patients in a cohort recruited in Durban, South Africa, during the first pandemic wave in July 2020 using detailed flow cytometry phenotyping of longitudinal samples with markers of B cell maturation, homing and regulatory features. ResultsThis revealed a coordinated B cell response to COVID-19 that differed significantly between HIV-ve and PLWH. Memory B cells in PLWH displayed evidence of reduced germinal center (GC) activity, homing capacity and class-switching responses, with increased PD-L1 expression, and decreased Tfh frequency. This was mirrored by increased extrafollicular (EF) activity, with dynamic changes in activated double negative (DN2) and activated naive B cells, which correlated with anti-RBD-titres in these individuals. An elevated SARS-CoV-2 specific EF response in PLWH was confirmed using viral spike and RBD bait proteins. ConclusionsDespite similar disease severity, these trends were highest in participants with uncontrolled HIV, implicating HIV in driving these changes. EF B cell responses are rapid but give rise to lower affinity antibodies, less durable long-term memory, and reduced capacity to adapt to new variants. Further work is needed to determine the long-term effects of HIV on SARS-CoV-2 immunity, particularly as new variants emerge. FundingThis work was supported by a grant from the Wellcome Trust to the Africa Health Research Institute (Wellcome Trust Strategic Core Award [grant number 201433/Z/16/Z]). Additional funding was received from the South African Department of Science and Innovation through the National Research Foundation (South African Research Chairs Initiative, [grant number 64809]), and the Victor Daitz Foundation.

immunology↗

bioRxiv2019

Contrasting inflammatory signatures in peripheral blood and bronchoalveolar cells reveal compartment-specific effects of HIV infection.

The mechanisms by which HIV increases susceptibility to tuberculosis and other respiratory infections are incompletely understood. We used transcriptomics of paired whole bronchoalveolar lavage (BAL) fluid and peripheral blood mononuclear cells to compare the effect of HIV at the lung mucosal surface and in the peripheral blood. The large majority of HIV-induced differentially expressed genes (DEGs) were specific to either the peripheral or lung mucosa compartments (1,307/1,404, 93%). Type I interferon signaling was the dominant signature of DEGs in HIV-positive blood with a less dominant and qualitatively distinct type I interferon gene set expression pattern in HIV-positive BAL. DEGs in the HIV-positive BAL were significantly enriched for infiltration with cytotoxic CD8+ T cells. Higher expression of representative transcripts and proteins in BAL CD8+ T cells during HIV infection, including IFNG (IFN-{gamma}), GZMB (Granzyme B) and PDCD1 (PD-1), was confirmed by cell-subset specific transcriptional analysis and flow cytometry. Thus, we report that a whole transcriptomic approach revealed qualitatively distinct effects of HIV in blood and bronchoalveolar compartments. Further work exploring the impact of distinct type I interferon programs and CD8+ T cells infiltration of the lung mucosa during HIV infection may provide novel insights into HIV-induced susceptibility to respiratory pathogens.

immunology↗