Biology subjects

Mukhopadhyay, S.

Publications and source records attributed to Mukhopadhyay, S..

At least 19 recordsLinked to original sources

bioRxiv2026

Topoisomerase IIβ Kinase: A protein processed from HIV-1 envelope by cysteine proteases

Topoisomerase II{beta} (TopoII{beta}) is activated by TopoII{beta} kinase (TopoII{beta}K) during HIV-1 infection. This study demonstrates that TopoII{beta}K is produced within 15-30 minutes post-infection and is conserved across the major HIV-1 subtypes (A, B, C, and D), although its formation kinetics vary among subtypes. Immunological mapping using specific anti-gp120 antibodies suggests that TopoII{beta}K is derived from the N-terminal region of gp120. Notably, TopoII{beta}K co-localizes with endolysosomes shortly after viral entry. Inhibition of cysteine proteases using E64d and Z-FY-CHO abrogates TopoII{beta}K production in both HIV-1-infected and envelope-transfected cells. These findings suggest that TopoII{beta}K is a conserved polypeptide product of gp120 processing within endolysosomes during the early stages of HIV-1 infection, highlighting a potentially conserved mechanism that regulates early viral infection across major HIV-1 subtypes.

molecular biology↗

bioRxiv2026

RNA m6A mediates de novo DNA methylation and heritable epigenetic silencing in human cells

DNA methylation is the most abundant epigenetic modification on the human genome and changes dynamically during embryogenesis, aging, and disease. How DNA methylation is established de novo remains incompletely understood. Using orthogonal CRISPR platforms to initiate de novo DNA methylation coupled with genome-wide CRISPR screens and proximity-dependent proteomics, we discover the interplay between RNA m6A modifications and de novo DNA methylation to program heritable epigenetic memory in human cells. The RNA m6A writer complex is recruited to chromatin with DNA methyltransferases and histone repressors during the transient establishment phase of heritable epigenetic silencing. Inhibition of RNA m6A during early pluripotent transition leads to defects in de novo DNA methylation, resulting in upregulation of transposable elements that pose a threat to genome integrity during early development. Our work reveals the vital coordination between the epitranscriptome and de novo DNA methylation to establish proper heritable epigenetic silencing in human cells.

genetics↗

bioRxiv2026

Global population genomics and pangenome of the clubroot pathogen link clonal expansion and effector diversification to resistance breakdown in canola

Clubroot disease, caused by the obligate biotrophic protist Plasmodiophora brassicae, threatens Brassica crops worldwide. Novel pathogen virulence frequently breaks resistance in agricultural crops, and yet explanations for isolate evolution remain unknown. We combined whole genome sequencing of 192 isolates from 24 countries with a pangenome analysis of 55 high-quality complete or nearly complete genome assemblies and performed functional effector analyses to investigate potential mechanisms of virulence emergence. Three major phylogenetic clusters were identified, including two highly clonal lineages with contrasting virulence profiles, and a genetically diverse population retaining signatures of recombination. Our findings support a model in which genome-wide purifying selection maintained core genome integrity, while effector genes evolved under less constraint. Despite a compact closed pangenome, we identified chromosome 17 as a major hotspot of single nucleotide and structural variations, transposable elements and expansion of ankyrin repeat effectors. Within the variable chromosome 17, PbAnk2 alleles from virulent and avirulent isolates encoded proteins with distinct predicted structures. PbAnk2 interacted with the Arabidopsis co-chaperone DJC31 in yeast two-hybrid and BiFC assays, whereas co-immunoprecipitation did not detect an interaction for the allele found in isolates that break CRa-mediated resistance, suggesting that allelic variation may alter PbAnk2's association with DJC31. These findings reveal the clonal expansion of virulent isolates and connect it to effector diversification that led to resistance breakdown and establish a genomic framework for pathogen surveillance and durable clubroot resistance.

genomics↗

bioRxiv2026

The chromosome level genome of the Blueberry Stem Gall Wasp, Hemadas nubilipennis (Hymenoptera: Ormyridae) on lowbush blueberry (Vaccinium angustifolium) reveals repeat-driven expansion

Gall-inducing wasps are emerging models for studying plantinsect coevolution, host manipulation, host plant adaptation, and speciation, yet chromosome-level resources remain scarce for most lineages. The blueberry stem gall wasp (BSGW), Hemadas nubilipennis (Hymenoptera: Ormyridae), is native to North America where it induces galls on both lowbush (Vaccinium angustifolium) and highbush blueberries (V. corymbosum). Recently, BSGW has reached outbreak densities in cultivated highbush production. Given that (a) the biology has been characterized primarily from natural lowbush-associated populations, (b) the absence of genomic resources limits comparative analyses, and (c) populations on cultivated highbush represent a recent host shift, we generated the first chromosome-level genome from wild lowbush blueberry. The BSGW genome consists of five chromosome-scale scaffolds totaling 1.08 Gb (N50 = 218 Mb), the second largest known in Chalcidoidea. Comparative analysis reveals that genome size variation is driven primarily by transposable element proliferation (R = 0.96, p < 0.001), with BSGW exhibiting a high proportion of unclassified TEs. Gene-body methylation is conserved, enriched in exons of broadly expressed core genes, and correlates with gene density. The mitochondrial genome (18,697 bp) exhibits extensive gene rearrangement, and COI sequences reveal 4.35.4% divergence from geographically distant populations, suggesting a complex of cryptic species. Additionally, we assemble a near-complete genome of the endosymbiont Wolbachia pipientis (Supergroup A), which encodes PifA and PifB effectors potentially linked to parthenogenesis. These resources establish a foundation for population genomics, taxonomic revision, and applied management, while providing insights into genome architecture, epigenetics, and symbiont interactions.

genomics↗

bioRxiv2026

Translocon Remodeling Modulates Ribosomal Frameshifting and the Maturation of the Sindbis Virus Structural Polyprotein

Like other enveloped RNA viruses, alphaviruses synthesize and assemble their envelope glycoproteins at the endoplasmic reticulum (ER) membrane. There, protein-conducting channels called translocons provide nascent proteins access to the membrane and allow them to fold into their correct shapes. We previously showed that a hydrophobic segment in the Sindbis virus structural polyprotein forms cotranslational interactions with the translocon that enhance -1 programmed ribosomal frameshifting (-1PRF), a recoding event that regulates polyprotein biogenesis. Recent discoveries concerning translocon remodeling suggest this segment, which corresponds to the second transmembrane domain of the E2 protein, could serve as a signal that recruits the multipass translocon (MPT). Here, we show that knocking out certain components of the MPT increases -1PRF efficiency. These differences in recoding coincide with changes in the membrane topology of the nascent polyprotein and in its downstream proteolytic processing in a manner that ultimately reduces viral fitness. Together, our results indicate that -1PRF and spike protein maturation in alphaviruses is tuned by the dynamic remodeling of the ER translocon. Such coupling could allow polyprotein biogenesis to adapt to different stages of viral replication and to distinct host or vector environments.

molecular biology↗

bioRxiv2026

RenSeq and whole genome sequencing uncover allelic diversity of clubroot resistance genes in commercial breeding canola lines

Clubroot disease, caused by the obligate biotrophic pathogen Plasmodiophora brassicae, is a major threat to canola (Brassica napus) production worldwide. Clubroot-resistant (CR) cultivars remain the most effective disease-management strategy, but the genetic basis of resistance in commercial canola remains poorly understood because many resistance sources are proprietary and associated genotypic information is rarely accessible. Although nucleotide-binding leucine-rich repeat (NLR) immune receptors account for most cloned CR genes, no pan-NLRome has incorporated CR lines used in commercial canola breeding. Here, we combined whole-genome sequencing and resistance gene enrichment sequencing (RenSeq) to assemble and annotate the NLR repertoires of five homozygous CR inbred lines (IH1-IH5) used for commercial breeding and displaying contrasting resistance profiles against predominant Canadian P. brassicae pathotypes. We integrated these NLRomes with the susceptible cultivar Westar to construct a comparative pan-NLRome for canola. Across the five CR lines, total NLR content was highly conserved, ranging from 504 to 517 genes, with TIR-NLRs representing the predominant class. C-JID-containing TIR-NLRs accounted for more than 30% of each NLR repertoire, and integrated-domain analysis identified conserved and genotype-specific NLR-IDs, including previously unreported domains in IH4. Pan-NLRome analysis resolved 366 NLR orthogroups (OGs), 60.7% of which were core, and identified resistant-line-enriched OGs absent from Westar as candidate CR-associated loci. Unexpectedly, a homolog of the functionally characterized CR gene, CRa, was detected in five CR lines. Moreover, a homolog of another CR gene, Crr1a, was detected in both resistant and susceptible lines, indicating that the presence/absence of a gene alone does not predict resistance. Instead, structural variation affecting LRR and C-JID regions suggests that allele-level diversity within conserved NLR loci contributes to CR-associated variation, with implications for allele-specific marker development and durable CR deployment.

plant biology↗

bioRxiv2026

The Aorta-Gonad-Mesonephros niche shapes the functions of yolk sac-derived macrophages involved in hematopoietic stem and progenitor cell generation ex vivo

Hematopoietic stem cells (HSCs) generated from induced pluripotent stem cells (iPSCs) offer a promising patient-specific alternative to allogeneic transplantation, yet current differentiation protocols fail to fully recapitulate in vivo HSC maturation. During mouse development, yolk sac (YS)-derived macrophages populate the aorta-gonad-mesonephros (AGM) region at the time of HSC emergence, but the mechanisms by which they support ex vivo hematopoietic stem and progenitor cell (HSPC) generation remain poorly defined. Bulk RNA sequencing revealed that mature AGM CD206 macrophages upregulate pro-inflammatory cytokines and the adhesion molecule F4/80. Using F4/80 knockout embryos, we identify a previously unreported, niche-specific role for F4/80 in restraining the frequency and colony-forming activity of HSPC subsets in the AGM, while supporting endothelial cell maintenance; this effect was absent in the YS. Lineage-tracing with a Cdh5-CreERT2;Rosa26LSL-tdTomato pulse-chase system confirmed that both CD206 and CD206- AGM cells originate from early YS-derived endothelial precursors, with no evidence of local macrophage generation within the AGM. Functional co-culture assays further demonstrated that the ability of CD206 macrophages to enhance the progenitor potential of hemogenic endothelium is AGM-specific and not an intrinsic, ontogeny-determined property, as YS macrophages failed to confer the same benefit even when paired with AGM endothelial cells, and AGM macrophages were ineffective with YS endothelium. Differential expression and NicheNet ligand-receptor interaction analyses identified a small set of AGM-restricted macrophage genes - including Mmp2, Nrep, Ccl2, and Cxcl16 - which are predicted to interact with both endothelial and cluster cells during endothelial-to-hematopoietic transition. Together, these findings establish that AGM macrophages acquire niche-specific transcriptional and functional properties upon entry into the aortic microenvironment, independent of their YS origin, and identify candidate macrophage-derived factors and a novel regulatory role for F4/80 in shaping HSPC output. These insights may guide the refinement of iPSC-based HSC differentiation protocols through the targeted, temporally controlled addition of macrophage-associated signals.

developmental biology↗

bioRxiv2026

Sequence-Based Therapeutic Peptide Classification with Augmented Negative Sampling

Therapeutic peptides offer high target specificity, low toxicity, and the ability to modulate protein-protein interactions, yet experimental functional characterization remains costly and slow. Computational prediction of therapeutic function directly from sequence could accelerate peptide screening and enable generative design pipelines, but requires reliable discrimination between therapeutic and non-therapeutic peptides. Existing multi-label predictors cover few functions, rely on limited datasets, and exhibit high False Positive Rates (FPRs), limiting their practical utility. We present a lightweight CNN classifier trained on the most comprehensive therapeutic peptide database to date (54,655 peptides, 48 functional categories). A key contribution is a statistically motivated negative sampling strategy using Markov models to generate diverse synthetic decoys at multiple difficulty levels. When evaluated on this controlled decoy benchmark, the FPR is reduced from over 60% for previous models to 2.1% for our approach. On positive therapeutic samples, our fine-tuned five-model ensemble achieves 79.9% Micro F1 and 54.6% Macro F1 while requiring only amino acid sequences as inputs. Analysis using a sparse L1-constrained variant of our model shows that convolutional filters capture conserved functional motifs and statistically improbable non-therapeutic patterns, with downstream layers combining these signals, providing mechanistic evidence that the network learns biologically meaningful structure. On an external generalization benchmark derived from TPpred-LE, our model achieves 55.3% Micro F1 and 38.6% Macro F1 on the 12 shared labels, close to the benchmark-specific baseline (57.9%/38.1%), while retaining substantially broader therapeutic label coverage. Code and models will be made available at https://github.com/terra-quantum-public/tq-therapep-ai.

bioinformatics↗

bioRxiv2026

An autoinhibitory regulatory switch governs heterotypic phase separation and liquid-to-solid phase transition of TDP-43 and tau into cytotoxic amyloids

Biomolecular condensates formed via macromolecular phase separation of proteins and nucleic acids control a myriad of essential cellular processes, whereas abnormal phase transitions into solid-like aggregates are associated with a range of fatal neurodegenerative diseases. Here, we present a unique case to demonstrate that two neuronal proteins, TDP-43 and tau, undergo heterotypic phase separation via domain-specific interactions regulated by an autoinhibitory conformational switch. Using single-molecule FRET (Forster resonance energy transfer), in combination with multi-color high-resolution imaging, fluorescence recovery after photobleaching, fluorescence correlation spectroscopy, single-droplet fluorescence anisotropy imaging, homoFRET microscopy, fluorescence lifetime-FRET imaging, vibrational Raman spectroscopy, and electron microscopy, we unmask the interplay of molecular drivers and dissect the sequence of events associated with the formation of TDP-43:tau co-condensates that undergo liquid-to-solid phase transitions into cytotoxic amyloids. Our cellular studies show that a disease-associated cytosolic fragment of TDP-43 recruits tau into oxidative stress-induced cytoplasmic granules, eliciting cellular toxicity. Our findings provide mechanistic underpinnings of co-condensation-mediated aberrant phase transitions associated with exacerbated neuropathological outcomes.

biophysics↗

bioRxiv2026

A clubroot pathogen PBS3-like effector manipulates hormonal crosstalk to alter root morphology during colonization

O_LISalicylic acid (SA) and auxin are key regulators of plant immunity and development. The clubroot pathogen Plasmodiophora brassicae encodes PbGH3, an effector related to the GH3 family involved in phytohormone homeostasis. Although PbGH3 was proposed to conjugate auxin in vitro, its biological function in planta has remained unclear. This study aimed to determine the in vivo role of PbGH3 during host colonization. C_LIO_LIWe generated Arabidopsis thaliana and Brassica napus lines overexpressing PbGH3 and characterized their developmental phenotypes, hormone profiles, gene expression, and susceptibility to infection. Structural modeling was performed to assess PbGH3 similarity to plant GH3 proteins, and functional complementation was tested using the Arabidopsis gh3.12 mutant. C_LIO_LIThe expression of PbGH3 in Arabidopsis induced auxin-related developmental phenotypes without detectable accumulation of auxin conjugates. Instead, PbGH3 structurally and functionally resembled GH3.12/PBS3 inducing increased conjugated SA levels, reduced jasmonic acid, suppressed PIN2 expression, and increased root hair number and infection. PbGH3 complemented SA-related defects in the gh3.12 mutant. C_LIO_LIPbGH3 functions as a modulator of SA metabolism rather than an auxin-conjugating enzyme, likely competing with host GH3.12/PBS3 to constrain effective SA accumulation. This reveals a novel strategy by which P. brassicae disrupts SA-auxin homeostasis to promote host colonization and ensure disease development. C_LI PLAIN LANGUAGE SUMMARYThis study shows that the clubroot pathogen uses a protein called PbGH3 to modify the plants salicylic acid balance. This alters root traits and increases susceptibility to infection. Arabidopsis and canola plants engineered to produce PbGH3 showed similar changes, revealing that the pathogen uses this protein to disrupt hormone regulation and create conditions that support its colonization.

plant biology↗

bioRxiv2026

A bacterial extracellular matrix protein forms a supramolecular metallogel

The microbial extracellular matrix (ECM) is a complex network of self-secreted biopolymers uniting the cells in biofilms, providing them with structural integrity, and contributing to their elevated resistance to antibiotic treatments. Recently, there is a growing realization that a regulated, bidirectional cross-talk of bacteria and ECM confers biofilms with tissue-like traits, however, the mechanisms of spatio-temporal self-organisation of ECM and its regulation are still poorly understood. In the model organism for biofilm formation Bacillus subtilis, TasA is the major protein component of the extracellular matrix. We recently showed that TasA, isolated in the form of stable and structured globules, assembles into elongated and ordered fibers via a donor-strand complementation mechanism. In this study, we discovered that in the presence of zinc metal ions, TasA is able to form hydrogels with > 97% water content. Electron- and atomic force-microscopies as well as small angle X-ray scattering measurements show that cross-linking with zinc ions induces a transition in TasA morphology from one-dimensional fibers to two-dimensional sheets. Electron paramagnetic resonance measurements then show that such a significant morphological shift is associated with molecular changes in the coordination environment of zinc ions, which lead to structural changes at the protein level. When assembling into macroscopic networks, TasA-Zn metallogels exhibit viscoelastic properties and a fast recovery following an excessive strain. These metallogels represent a novel class of bacterially-derived ECMs that form easily at room temperature without covalent crosslinking, and may be used as a natural matrix-mimics in biofilm models for infection studies.

biochemistry↗

bioRxiv2026

Strategic Segmentation: A Nash Equilibrium based Approach for Weed Segmentation in Agricultural Fields

Weeds present a major challenge to agricultural productivity, competing with crops for critical resources like water, nutrients, and sunlight, resulting in significant yield reductions. Prompt weed identification is essential for enabling effective control strategies, such as the application of herbicides or mechanical removal, to minimize their impact on crop growth. This research focuses on developing a deep learning approach based on game theory for detecting weeds. Using CWFID dataset captured at various times and days, along with multispectral data in the visible and near-infrared spectrum, the study aims to improve early detection methods for more efficient weed management in agricultural settings. A novel segmentation technique for weed regions is introduced, employing a zero-sum game theory model to reconcile conflicting classifications from different weed detectors. These regions are treated as zones of conflict between weeds and crops, with each detector representing a different strategy. By defining an appropriate utility function, the method identifies the Nash equilibrium, effectively minimizing false positive detections of weeds.

plant biology↗

bioRxiv2025

The regulation of virulence gene expression is controlled by phase separation of heterochromatin protein 1 (HP1) in Plasmodium falciparum

Malaria caused by Plasmodium falciparum involves antigenic variation on the infected red blood cell surface by a mutually exclusive expression of virulence (var) genes. The repressed var genes have a distinct genome organization where they localize as a cluster near the nuclear periphery and are bound by PfHP1 (P. falciparum Heterochromatin Protein 1). However, the mode of regulation of var genes by PfHP1 remains unclear. In this study, we show that PfHP1 undergoes phase separation in vitro in an RNA and DNA-dependent manner. Single-molecule DNA tethering experiments further revealed that AT-rich DNA sequences act as nucleation sites for the assembly and compaction of PfHP1-mediated heterochromatinization. We have also identified point mutations in the IDRs (intrinsically disordered regions) of PfHP1 that disrupt its phase separation as well as DNA compaction in vitro. To assess the dynamic properties of PfHP1 condensates in vivo, we performed fluorescence recovery after photobleaching (FRAP) in GFP tagged PfHP1 parasites, which revealed rapid fluorescence recovery, supporting their fluidity and phase separation behaviour. Ectopic expression of PfHP1 phase separation and DNA compaction mutants led to dispersed nuclear localization of PfHP1, in contrast to the punctate appearance of the wild-type protein and altered chromatin binding at var genes. These results were corroborated with DiCre/loxP-based conditional expression of the same PfHP1 mutants, which also led to the de-repression of multiple var genes (as many as 54 out of 60 var genes), mimicking the phenotype of PfHP1 depletion mutants. Hi-C sequencing of PfHP1 mutants revealed loss of interactions in the heterochromatic regions, indicating that PfHP1 phase separation is an essential mechanism for repressive cluster formation. In conclusion, our study demonstrates the role of PfHP1-mediated phase separation in heterochromatin formation and var gene silencing; unveiling a fundamental mechanism that drives antigenic variation in P. falciparum.

molecular biology↗

bioRxiv2025

L2 loop engineering enhances the enzymatic activity and synergism property of the AA16 family Lytic polysaccharide monooxygenase enzyme from Aspergillus fumigatus.

Lytic polysaccharide monooxygenase (LPMO) is an enzyme that has enormous potential for industrial applications. It has a synergistic effect with the cellulase enzyme complex. The LPMO enzymes typically adopt a compact {beta}-sandwich fold that consists of a 7 to 9 {beta}-strand with a flat active site containing copper in its active site. There are a few loops in LPMO, among them the L2 loop is reported to take a key role in shaping the active site and substrate binding. Now, in this work, we want to investigate the role of the L2 loop in enzymatic activity and synergistic effect. In achieving our goal, we have replaced the L2 loop of our concerning AfLPMO16 with other L2 loops from different LPMOs: HiLPMO9B (PDB: 5NNS), McLPMO9 (PDB: 7NTL) and CsLPMO9 (PDB: 7EXK). Interestingly, L2 loop replacement from CsLPMO9 (PDB: 7EXK) showed enhanced activity and synergism compared to others. The secondary structural analysis by circular dichroism also suggested that it changed the structure significantly. Moreover, this is the first report of complete L2 loop engineering in LPMO.

biochemistry↗

bioRxiv2025

The PPE2 protein of Mycobacterium tuberculosis is responsible for the development of hyperglycemia and insulin resistance during tuberculosis

Diabetes is a known risk factor for tuberculosis (TB), but clinical evidences suggest that TB itself can induce hyperglycaemia and insulin resistance, though the underlying mycobacterial factors are not known. Herein, we implicate PPE2, a secretory PE/PPE family protein of Mycobacterium tuberculosis (Mtb), as a key modulator of adipose tissue physiology that contributes to the development of insulin resistance. In mice, PPE2 caused fat loss, adipocyte hypertrophy, immune cell infiltration, impaired glucose tolerance, reduced expression of PPAR-{gamma}, C/EBP-, adiponectin and higher insulin resistance. Transcriptomic analysis revealed PPE2 altered expression of genes associated with chemokine/cytokine, ribosomal biogenesis and lipase signaling. PPE2 induced lipolysis by activating cAMP-PKA-HSL axis, increased circulating free fatty acids, a feature also observed in TB patient sera. Interestingly, PPE2-immunization mitigated these effects, suggesting its potential as a subunit vaccine. Overall, this study identifies PPE2 as a key link between Mtb-infection, adipose tissue dysfunction and insulin resistance.

microbiology↗

bioRxiv2025

ESAT-6 protein of Mycobacterium tuberculosis inhibits differentiation of human monocytes to dendritic cells

Mycobacterium tuberculosis (Mtb) employs multiple strategies to evade host immunity, including disruption of antigen presentation. Dendritic cells (DCs) are crucial for effective antigen presentation and T-cell activation. In this study, we show that mycobacterial protein, ESAT-6, impairs monocyte to DC differentiation with reduced expression of the DC marker, CD209. ESAT-6 treatment elevated IL-6 and IL-10 levels, but blocking of these cytokines failed to restore DC differentiation. Mechanistically, ESAT-6 suppressed phosphorylation of p65, establishing that ESAT-6 impairs DC differentiation by inhibiting NF-{kappa}B activation and this function is dependent on the last six amino acids of its C-terminal domain. This mechanism may represent a novel immune evasion strategy employed by Mtb to subvert host adaptive immune responses during infection.

immunology↗

bioRxiv2025

PE11 promotes intracellular persistence of Mycobacterium tuberculosis by inhibiting autophagy and lysosomal biogenesis by targeting the FLCN-lactate-TFEB signaling axis

Mycobacterium tuberculosis (Mtb) employs multiple virulence factors, including cell wall-associated proteins, to evade host immune responses. PE11, a cell wall-localized esterase, contributes to Mtb persistence by facilitating cell wall remodelling and resistance to acidic and antibiotic stress. Herein we describe a novel role of PE11 in subverting host autophagy through disruption of TFEB-mediated lysosomal function. PE11 promotes FLCN-dependent depletion of intracellular lactate to destabilize TFEB and thereby downregulate genes essential for autophagic flux and lysosomal acidification. Using a PE11-deficient Mtb strain, we demonstrate that PE11 targets the FLCN-lactate axis to regulate TFEB stability. Exogenous lactate supplementation restored TFEB stability, enhanced lysosomal acidification, and significantly reduced intracellular bacterial burden. Lactate also synergized with frontline anti-tubercular drugs to improve Mtb clearance. These findings establish PE11 as a key immune evasion factor and highlight lactate as a promising host-directed therapeutic to enhance bacterial killing and reduce antibiotic-associated toxicity.

microbiology↗

bioRxiv2025

Multi-Model LLM Architectures for Personalized Summarization and Relevance Ranking in Biomedical Literature

ObjectiveTo develop and evaluate a personalized literature review system that efficiently processes and summarizes biomedical literature to provide timely, relevant insights for researchers. MethodsThe system integrates ontology-aware keyword extraction (MeSH/ACM constrained TF-IDF from CV/Research Statement), citation-informed retrieval (PubMed and NIH iCite API), and dual-model large language model (LLM) summarization (Google Gemini 2.0 flash, OpenAI GPT-4o-mini). These LLMs leverage advanced Transformer architectures, building on foundations such as BERT, BART, and BioBERT. A two-stage ranking algorithm combines Relative Citation Ratio (RCR) with cosine similarity. Summary quality was evaluated using ROGUE-1/2/L and BERTScore. The system is deployed as a Streamlit web application. ResultsAcross 20 biomedical queries, the system demonstrated strong average performance (BERT-F1{approx} 0.86), with cosine similarity strongly correlating with summary quality. Human evaluation involving 10 users yielded average scores above 4.5/5 across summary fidelity and keyword relevance. ConclusionHybrid ranking and ensemble LLM summarization significantly accelerate scientific sense-making. These findings suggest broad applicability to various domains beyond biomedicine.

bioinformatics↗