Integrative transcriptomics and hypothesis-driven transfer machine learning reveal conserved and species-specific host-parasite dynamics across Leishmania species in THP-1 cells: a systematic review and meta-analysis.

INTRODUCTION: Leishmaniasis is a vector-borne parasitic disease caused by protozoa of the genus Leishmania, characterized by clinical outcomes ranging from self-limiting cutaneous lesions to fatal visceral disease. Disease progression is multifactorial and is largely shaped by interactions between the parasite and host macrophages, as well as by other host- and pathogen-related factors. Although transcriptomic studies have provided insights into these interactions, differences in experimental design, parasite species, and analytical workflows have limited cross-study comparisons and the identification of conserved molecular responses. METHODS: A systematic review was conducted following PRISMA guidelines to identify publicly available RNA-seq datasets of Leishmania-infected THP-1 macrophages. Raw sequencing data from eligible studies were reanalyzed using a unified bioinformatics pipeline incorporating standardized quality control, differential gene expression analysis, and batch-effect correction to enable robust cross-study integration. Comparative analyses were performed across infection stages and between L. infantum and L. amazonensis. The integrated transcriptomic dataset was subsequently used to develop a hypothesis-driven transfer learning framework to evaluate the feasibility of predicting L. amazonensis parasite gene expression at 96 hours post-infection (hpi) from experimentally generated 24 hpi transcriptomic profiles. RESULTS: Integrated analysis revealed a pronounced early induction of pro-inflammatory and interferon-stimulated genes, including CXCL10, IL1B, and IFIT1, followed by attenuation of inflammatory signalling at later infection stages. Comparative analyses identified a conserved interferon-driven host response shared between L. infantum and L. amazonensis, together with species-specific transcriptional adaptations. The transfer learning framework demonstrated the feasibility of predicting late-stage parasite gene expression from early transcriptomic data, highlighting the potential of machine learning approaches to leverage limited transcriptomic datasets. DISCUSSION: This study provides an integrated transcriptomic framework for investigating host-parasite interactions in Leishmania-infected macrophages and identifies conserved and species-specific transcriptional responses across infection. Furthermore, the proposed hypothesis-driven transfer learning approach demonstrates the potential to address transcriptomic data scarcity in neglected tropical disease research. Future in vitro studies and the availability of additional transcriptomic datasets will facilitate improved model training, validation, and generalizability.