Science China Life Sciences

MotivationGut microbiota regulates host health through complex protein-protein interactions. However, deciphering this specific interactions between microbiota and human receptors remains a significant challenge due to the lack of specialized computational tools. ResultsLeveraging the hypothesis of cell communication and relevant data, HMI-Pred initially builds an ensemble classifier to screen for potential ligand sequences within microbial genomes. It then jointly evaluates sequence semantics and molecular docking to predict potential microbe-host receptor interactions.HMI-Pred achieved robust performance with F1-scores of 0.901 for microbial ligand identification and 0.883 for interaction prediction. Application to 332,381 microbial proteins revealed distinct interaction patterns: histone deacetylases (HDACs) served as broad-spectrum targets (mean score > 0.80), while G protein-coupled receptors (GPCRs) exhibited high specificity (scores 0.42-0.61). Furthermore, literature mining validated over 47% of the functional predictions, and specific immunomodulatory interactions were confirmed in Akkermansia muciniphila.HMI-Pred provides a valuable computational tool for decoding host-microbe signaling networks and facilitating the discovery of microbiome-based therapeutic targets. AvailabilityThe source code and documentation are available at https://github.com/YangLab-BUPT/HMI-Pred. Contactlihm@bupt.edu.cn

Optogenetic neuropathway inhibition is a powerful approach for dissecting circuit functions. This strategy, however, frequently encounters practical challenges due to insufficient expression or performance of the optogenetic silencer on axonal projections/terminals. HcKCR1, a light-gated potassium-selective channel from Hyphochytrium catenoides, has shown great promise for optogenetic inhibition. Unfortunately, the application of HcKCR1 in neuropathway manipulations is hindered by its unsatisfactory gating properties and poor axonal trafficking. To overcome these hurdles, we first engineered a performance-improved HcKCR1 (piKCR) that allowed more reliable neuronal inhibition at low intensities of green or red light. We next engineered an axon-targeted piKCR (piKCR.AT) that demonstrated long-range axonal trafficking and optical presynaptic inhibition in the mouse hippocampus. When piKCR.AT was expressed in the cerebellar Purkinje Cells (PCs), optical manipulation of PC outputs to the deep cerebellar nuclei robustly disrupted mouse movement on the balance beam. With enhanced performance and axonal distribution, piKCR.AT may provide new opportunities for elucidating neuropathway functions in health and diseases.

Wheat grain weight and flour quality largely depend on starch biosynthesis, yet the mechanisms by which transcription factors coordinate this process remain poorly understood. In this study, using an integrative strategy that combines genome-wide association analysis with yeast two-hybrid library screening, we identify TaNF-YC10, a Nuclear Factor Y transcription factor, as a positive regulator of starch accumulation in the wheat endosperm. Loss of TaNF-YC10 reduces starch content and alters starch granule size distribution, whereas overexpression enhances starch accumulation and increases grain weight. TaNF-YC10 binds and activates core starch biosynthetic-related genes, including AGPL1, GBSS1, YUC11, and NF-YB7, and forms higher-order transcriptional complexes with TaNF-YB1 and TabHLH95 to coordinate multiple regulatory pathways. TaNF-YC10-A1-Hap2 is associated with higher starch content and thousand grain weight and has been selected during wheat breeding in China. Collectively, our findings establish TaNF-YC10 as a pivotal transcriptional hub in starch regulation and highlight its potential as a target for genetic improvement of grain yield in wheat.

The newly identified signaling system comprising C-X-C motif chemokine ligand 17 (CXCL17) and G protein-coupled receptor 25 (GPR25) is involved in immune regulation and tumor development. However, the evolutionary origin of this pair has remained unclear because CXCL17 orthologs in lower vertebrates exhibit extreme sequence variation and cannot be identified through conventional homology-based searches. In this study, we identified seven possible CXCL17 orthologs in primitive cartilaginous fishes, including sharks and rays, using an integrated approach based on key amino acid sequence features as well as gene synteny, architecture, and RNA sequencing data in the NCBI gene database. To validate these candidates, a representative ortholog from the cloudy catshark (Scyliorhinus torazame), termed St-CXCL17, was prepared via bacterial overexpression and in vitro refolding. In cell-based functional assays, St-CXCL17 demonstrated high binding affinity and activation potency toward its corresponding receptor, St-GPR25. Further analysis revealed that removing three conserved C-terminal residues almost completely abolished this activity. While these cartilaginous fish CXCL17s share considerable homology with one another, they lack significant overall similarity to orthologs in mammals, amphibians, or bony fishes. These findings identify functional CXCL17 orthologs in cartilaginous fishes for the first time, implying that the CXCL17-GPR25 signaling pair likely originated in ancient cartilaginous fish ancestors or earlier and has been conserved throughout the evolution of jawed vertebrates.

Geothermal springs, characterized by extreme physicochemical conditions, represent ecologically and evolutionarily significant habitats that foster unique microbial communities and drive adaptive evolutionary processes. Despite their importance, the complex microbial interactions and underlying mechanisms governing community assembly in these environments remain poorly understood. In this study, we conducted systematic sampling across 49 geothermal springs in Tengchong, Yunnan, over a six-year period (2016-2021), and performed metagenomic sequencing on 152 samples. We successfully reconstructed 12,789 non-redundant microbial genomes, revealing an exceptionally high level of phylogenetic and functional diversity within the spring microbiomes. Our analyses demonstrate that pH and temperature are the primary deterministic drivers shaping both microbial species composition and functional potential, thereby segregating the communities into three distinct groups: acidic, hyperthermal, and thermal. Furthermore, ecological network analysis revealed that extreme environmental conditions significantly alter network topology, resulting in less complex but more efficient microbial interaction networks. Collectively, this study provides a comprehensive resource and mechanistic insights into the microbial diversity, community structure, and species interactions in geothermal spring ecosystems.

Enhancer-regulating epigenetic modifiers play critical roles in normal physiological processes and human pathogenesis. The major enhancer regulator paralogs MLL3 and MLL4 (MLL3/4) belong to the lysine methyltransferase 2 (KMT2) family, which catalyzes the methylation of lysine 4 on histone H3 (H3K4me). MLL3/4 are required for enhancer activation and are essential for mammalian development and stem cell differentiation. Recent studies have linked MLL3/4 with different metabolic pathways in the context of stem cell self-renewal and cancer cell growth; however, the underlying mechanisms remain elusive. Here, we utilize Seahorse extracellular flux analysis, stable isotope tracing, stem cell biology techniques, and transcriptomic analysis to investigate the functional relationship of MLL3/4, cellular respiration, and stem cell differentiation. Our results indicate that the loss of MLL3/4 impairs glycolytic activity and mitochondrial respiration in murine embryonic stem cells by downregulating the rate-limiting glycolytic enzyme Hexokinase 2 (HK2) and impairing the function of the Alpha-ketoglutarate dehydrogenase (OGDH) complex. Furthermore, simultaneously overexpression of HK2 and OGDH rescues defects in both cellular respiration and differentiation caused by MLL3/4 loss. Taken together, our study reveals a novel mechanism by which epigenetic machineries such as MLL3/4 govern the differentiation of pluripotent stem cells and facilitates the understanding of disease pathogenesis driven by enhancer malfunction.

Whole-genome duplication (WGD) is a major evolutionary event that drives molecular and species diversification. However, few studies have traced how WGD has shaped the long-term functional evolution of individual genes. Here, we investigated the olfactory marker protein (omp) genes duplicated by the teleost-specific WGD ([~]300 million years ago) through phylogenetic, syntenic, expression, and promoter analyses. Our results suggest that the duplicated omp gene pair has retained redundancy over an extended evolutionary period, leading to both non- and sub-functionalization, thereby generating molecular diversity. Moreover, evolutionary analyses of the olfactory signal transduction cascade revealed prolonged redundancy across its components, likely constrained by gene dosage balance. These findings imply that WGD may have introduced unexpected diversity into the entire olfactory signaling machinery of teleosts through dosage-constrained functional divergence. HighlightsO_LITeleost-specific WGD generated duplicated omp genes that persisted for [~]300 million years. C_LIO_LIExtended redundancy in ompa/ompb led to both non- and sub-functionalization. C_LIO_LIOlfactory transduction genes also show long-term redundancy shaped by dosage constraints. C_LIO_LIWGD likely introduced diversification into teleost olfactory signaling via dosage constraints. C_LI

ABSTRACPorcine respiratory diseases caused by extraintestinal pathogenic Escherichia coli (ExPEC) pose a severe threat to swine production and public health; however, research on respiratory tract-isolated ExPEC remains limited. This study comprehensively analyzed the genomic characteristics and antibiotic resistance gene (ARG) transfer potential of 441 ExPEC strains isolated from porcine lungs across 21 Chinese provinces (including 53 newly isolated strains from 2022-2024 and 388 NCBI-deposited strains). Phylogenetic analysis revealed that 84% of the isolates belonged to phylogroups A, B1, and C, with ST410, ST101, and ST88 as the predominant STs. The strains exhibited extensive ARG diversity, harboring 111 distinct ARG subtypes, with sul2 (81.4%), floR (73.5%), and tet (A) (68.0%) being the most prevalent. Importantly, critical "last-resort" antibiotic resistance genes (e.g., blaNDM-1/5, the mcr family, and tet (X4)) were also detected. Notably, 77.2% of the ARGs presented horizontal transfer potential, with plasmids (especially IncF family replicons) serving as core vectors, followed by integrons and transposons. Cooccurrence network analysis identified aph (3)-Ib, aph (6)-Id, sul2, and floR as core subnetworks driving multidrug resistance dissemination. Pangenomic analysis confirmed an open genome architecture, with core genes accounting for only 6%, reflecting the strains capacity to acquire exogenous genetic material via horizontal transfer. From the One Health perspective, these transferable ARGs can spread to the environment and humans through fecal discharge and the food chain. These findings underscore the importance of monitoring and controlling ExPEC infections in swine, as such strains can as reservoirs of ARGs, pose potential risks to human health, and may even act as sources of pathogenic agents responsible for human infections. IMPORTANCEPorcine respiratory ExPEC-induced diseases threaten swine production and public health, yet respiratory tract-isolated ExPEC research remains scarce. This study comprehensively analyzed 441 porcine lung ExPEC strains across 21 Chinese provinces, uncovering their dominant phylogroups, high ARG diversity (111 subtypes) and the presence of "last-resort" antibiotic resistance genes. We identified 77.2% of ARGs with horizontal transfer potential, plasmids (especially IncF family) as core vectors, and a core ARG subnetwork driving multidrug resistance. The open pangenome (6% core genes) highlights ExPECs strong capacity to acquire exogenous genes. These findings fill the research gap of respiratory ExPEC, clarify ARG transmission mechanisms in swine ExPEC, and provide critical genomic data for One Health-based AMR surveillance and control, guiding targeted strategies to mitigate ARG spread from swine to humans and the environment.

Live imaging is one of the most powerful methods to reveal the morphogenesis of plant organs. However, the highly three-dimensional structure of plant organs always poses technical challenges. For example, the basal region of leaf primordia is rarely observed because of the shape of leaf primordia and the sudden shift in geometry at the point where the leaf primordium connects to the hypocotyl. In this work, we developed a new live-imaging system that is suitable for observing the developmental process of the basal region of Arabidopsis leaf primordia at early stages. Using this system, we achieved continuous observation of the basal region of early Arabidopsis leaf primordia for more than 50 hours.

Somatic cell nuclear transfer (SCNT) holds great promise for regenerative medicine and agriculture, but its application is severely hampered by low efficiency, primarily attributable to aberrant epigenetic reprogramming. Although embryonic stem cells (ESCs) and trophoblast stem cells (TSCs) have been successfully derived from cloned embryos, an in vitro counterpart of the primitive endoderm (PrE) lineage has remained unavailable. To address this gap, this study reports the first successful establishment of extra-embryonic endoderm stem cell lines (XENs) from mouse SCNT-derived blastocysts (NT-XENs). Under conventional culture conditions, NT-XENs were generated from hybrid B6D2F1 blastocysts at a high efficiency of 55%, comparable to that of fertilization-derived XEN lines (FD-XENs, 50%), whereas derivation from inbred C57BL/6J SCNT-derived blastocysts was markedly lower (12.5%). Immunofluorescence and NanoString multiplex gene expression profiling confirmed that NT-XENs robustly expressed specific marker genes for PrE/XENs (e.g., Gata4, Gata6, Sox17), while exhibiting negligible or absent expression of pluripotency and trophoblast markers. Based on NanoString assay data, NT-XENs and FD-XENs shared highly similar global gene expression patterns, yet also exhibited some nonnegligible differences, exemplified by the differentially expressed genes (DEGs) Pecam1, Gtl2, Thbd and Xlr3b, which may suggest that the NT-XENs resided in a more differentiated state (potentially biased toward parietal endoderm (PE)) and retained SCNT-specific epigenetic imprinting errors, including aberrant X-chromosome inactivation and dysregulation of imprinted domains. In summary, this study successfully establishes NT-XEN cell lines, providing a valuable in vitro model for investigating the reprogramming scenarios of PrE lineage in SCNT and offering novel insights into the mechanisms underlying developmental failure of cloned embryos.

Characterized by the earliest use of pottery, the Jomon culture was a unique Neolithic culture that spread throughout the Japanese Archipelago. Previous archaeological evidence suggests that Jomon hunter-gatherers colonized the southernmost islands, the Ryukyu Archipelago, by approximately 7,000 years before present (YBP). However, genetic characteristics of the Ryukyu Jomon population and its contribution to the modern population have not been elucidated yet. In this study, we newly sequenced 273 modern and 25 ancient (6,700-900 YBP) whole genomes collected across the Ryukyu Archipelago. Our analysis demonstrated a genetic differentiation between the Hondo (Japanese mainland) and Ryukyu Jomon, dating back to [~]6,900 YBP. After the divergence from the Hondo Jomon, the Ryukyu Jomon experienced severe bottlenecks, with an effective population size of [~]2,000. Admixture between the Ryukyu Jomon and migrants from the historic Hondo population occurred [~]1,000 YBP, which corresponds to the widespread adoption of iron tools and agriculture in the Central Ryukyus. Different demographic histories between modern Hondo and Ryukyu populations resulted in different rates of Jomon ancestry in these populations. By providing a new perspective on the peopling of the Ryukyu Archipelago, this study significantly enhances our understanding of cultural transitions in the region.

Sandy beaches are dynamic coastal interfaces shaped by strong physical forcing and intense exchange between marine and terrestrial environments, yet their microbiomes remain poorly resolved at the genomic scale. Here we present a genome-resolved survey of microbial and viral communities across sandy beaches spanning a continental-scale latitudinal gradient along the Chinese coastline. By integrating cross-shore sampling, coastal geochemistry and large-scale multi-omics, we generated 978 metagenomes, 63 viromes and 72 metatranscriptomes, reconstructing 13,337 metagenome-assembled genomes and 38,255 viral populations. Sandy beach microbiomes exhibit exceptionally high genomic novelty, with more than 90% of species-level genomes representing previously undescribed taxa, suggesting that permeable coastal sediments constitute a distinct microbial and viral reservoir. Tidal zonation emerged as a dominant ecological driver structuring microbial diversity, metabolic strategies and virus-host interactions across cross-shore gradients. Genome-resolved analyses revealed systematic metabolic shifts from oxic heterotrophy in supratidal sediments toward increasingly chemolithotrophic and autotrophic pathways toward the low-intertidal and subtidal zone. Sandy beach microbiomes further encode broad potential for hydrocarbon and plastic transformation, together with diverse biosynthetic and antibiotic resistance repertoires that may mediate microbial chemical interactions. Together, these findings identify sandy beaches as a previously under-recognized microbial-viral biome shaped by tidal forcing, providing insight into microbiome evolution and coastal ecosystem resilience under increasing anthropogenic pressure.

Droplet sorting technology has the potential to revolutionize the biotechnology sector as it provides massive high-throughput screening capacity, but the technology remains not accessible for a wider audience yet. There is a need for more affordable droplet sorting platforms to design cell factories and screen cell libraries. In here we demonstrate our droplet cytometry/sorter platform for single-cell screening of yeast cells based on their fluorescence signal.

High and low tides occur twice a day (every [~]12.4 hours), with the largest tidal ranges during spring tides around new and full moons (every [~]14.765 days). While these lunar cycles are known to influence many animal phenotypes, particularly the reproduction of coastal animals, the genetic basis of lunar-related rhythms remains unclear. Since phenotypic variation is a valuable resource for elucidating such mechanisms, we examined geographic variation in the lunar-regulated mass spawning of the grass puffer (Takifugu alboplumbeus) along the Japanese coast. We found that western populations spawn during the first half of the spring tides, whereas eastern populations spawn during the second half. Furthermore, although spawning typically occurs a few hours before high tide, this timing is restricted to a specific time window that is earlier in the western populations than in the eastern ones. Behavioral analysis of larvae also revealed a shorter free-running circadian period ({tau}) in the western population than in the eastern ones. As differences in {tau} affect individual variation in the timing of physiological functions and behaviors, we hypothesized that differences in {tau} could account for the different time windows and consequently the observed difference in spawning days. Population genomics analysis identified proline-rich transmembrane protein 1-like (prrt1l) as a candidate gene. Expression of prrt1l was observed in the circadian pacemaker suprachiasmatic nucleus, and triple CRISPR F0 knockout of prrt1l shortened the free-running period in larvae. These findings suggest a potential mechanism underlying the geographic variation in lunar-synchronized spawning behavior. HighlightsO_LIThe geographic variation exists in the lunar-regulated spawning of the grass puffer, with differences in spawning dates and times between western and eastern Japan. C_LIO_LIThe free-running period of western populations is shorter than that of eastern populations, which is consistent with their earlier spawning timing. C_LIO_LIPopulation genomics analysis identified prrt1l as a candidate gene harboring population-specific missense mutations, the knockout of which shortens the free-running period. C_LI

Antibiotic resistance is a major global health concern. The development of new antibiotics and therapeutics is crucial for the future. Bacteriophages produce endolysins that induce bacterial lysis, making them a promising treatment option. Deep sequencing was used to identify and isolate genes encoding endolysins from Bacillus spp.. We characterized the biological activities of these endolysins. Additionally, this study focused on the design of a new chimeric endolysin, ArtE2, which combines endolysin-2 with a polycationic peptide to address the bacterial activity in gram-negative pathogenic bacteria. Using bioinformatic tools, we conducted three-dimensional modeling of the endolysin ArtE2 and its interactions with peptidoglycan fragments. In this study, we tested the activity of chimeric endolysins against both Gram-positive and Gram-negative bacteria. All endolysins share the same catalytic domain and diverse cell-binding domains. Some endolysins are highly specific to certain bacterial species or strains, whereas others have broader specificities. Histidine interactions are an important part of the mechanism by which ArtE2 connects with bacterial peptidoglycan. Additionally, the engineered endolysin ArtE2 was highly effective at killing Staphylococcus aureus and Escherichia coli. In silico analysis showed that the fusion did not negatively affect endolysin folding or activity. These findings suggest that ArtE2 could be used to develop efficient antibacterial controls targeting pathogenic Gram-positive and Gram-negative bacteria.

Cell mechanics can serve as an important biomarker for cell state and phenotype, such as metastatic ability. While some molecular mechanisms underlying cell mechanical properties have been investigated through targeted analyses, a genome-wide study of human genes and gene networks that modulate cell biophysical properties has not been attempted. In this work, we combined a microfluidic stiffness-based sorting device with a genome-scale CRISPR knockout (GeCKO) screen in order to investigate the effect of individual gene knockouts on cell stiffening and cell softening across the entire protein-coding genome. We processed approximately 150 million Cas9-expressing ovarian cancer cells that had been transduced with a library of 76,000 single guide RNAs (sgRNAs) against the 19,000 protein-coding genes in the genome. The cells were sorted into 5 mechanical subsets. We identified 7 gene knockouts that were significantly depleted in the softer subsets and over 700 gene knockouts that were significantly enriched in the stiffer subsets. Of these significant genes of interest, we selected 3 genes that were highly expressed in our ovarian cancer cell line with greater than 100-fold enrichment in the stiff outlet and resulted in significant changes in ovarian cancer patient survival. These genes, PIK3R4, CCDC88A, and GSK3B, when knocked out result in a significant and predicted increase in cell stiffness. This study is the first to explore the relation between human gene expression and cell mechanics at the genome-scale to generate datasets at the intersection between cell genotype, mechanotype, and phenotype for metastatic cancer cells. The method could also be applied to study the effect of genes on other biophysical cell processes as well as for identifying pathways for the control of cellular mechanics across many cell types.

Increasing attention is being focused on the glycemic index (GI) of daily food for humans, and the resistant starch content (RSC) is an important indicator of GI for starch-rich staple foods. In recent years, some studies revealed that the loss function of single or multiple key enzymes in the primary pathway of starch synthesis substantially increases RSC in rice, such as starch branching enzyme IIb (BEIIb) and soluble starch synthase IIIa (OsSSIIIa). However, a noteworthy negative characteristic of these high RSC mutants is the substantially increased amylose content (AC). AC as a major determinator of rice eating quality, must not be higher than an acceptable limit for most consumers. To solve this problem, in this study, we adopted two promoter editing (PE) editing strategies to develop rice germplasms with a better balance of RSC and AC: one is to edit the promoter of BEIIb in a low AC rice variety, another is to edit the promoter of Waxy (Wx) gene in a BEIIb loss of function mutant. Using AC[≤]20%, which is the range of premium quality rice in China as a criteria, we finally obtained 2 homozygous lines with significantly increased RSC ([≥]5%) in the NG46 background by promoter editing of BEIIb and 1 homozygous line in the YouTang2 (YT2, a BEIIb mutant) background by promoter editing of Wx gene. Further analysis revealed that AC and the amount of long-chain branches of amylopectin are positively correlated with RSC in the population of BEIIb PE lines. However, unexpectedly, the Wx PE-line with substantially decreased AC (17.7%) also showed significantly increased RSC (16.9%). Our study not only produces useful germplasms for the high RSC rice breeding in the future but also provides an insight into understanding the relationship between AC and RSC in defective BEIIb rice.

The intimate cohabitation between humans and their pets facilitates bidirectional microbial exchange, yet the extent and functional consequences of this transfer within the oral niche remain underexplored. Here, we employed metagenomic sequencing to characterize the oral microbiome of dogs and their owners across distinct geographic regions in China, integrating taxonomic, gene-centric, and functional analyses using public databases (BacMet, CARD, eggNOG, KEGG) to assess microbe-host associations. We found that dog-owner pairs exhibited significantly higher gene-level similarity compared to unrelated individuals, indicating a strong cohabitation-driven microbial linkage. While no major taxonomic shifts were observed in the human oral microbiome associated with pet ownership, we identified a marked enrichment of antibiotic resistance genes (ARGs)--particularly those conferring resistance to peptides, fluoroquinolones, antiseptics, diaminopyrimidines, cephalosporins, and carbapenems--in cohabiting pairs. This enrichment, together with the identification of exclusively shared ARGs (e.g., mdtF, macB, RanA), suggests the potential for horizontal gene transfer (HGT) between pet- and human-associated microbiomes. Functional profiling further revealed greater similarity in microbial metabolic pathways between cohabiting pairs than between unrelated individuals, reinforcing the likelihood of HGT as a mechanism underlying functional convergence. Collectively, these findings reveal that cohabitation with dogs reshapes the human oral microbiome at the genetic and functional levels, with potential implications for antimicrobial resistance transmission. This study provides a foundational framework for assessing the health risks associated with pet-human microbial exchange in shared living environments.

Cultured meat technology, with its significant advantages of shortening meat production cycles, reducing natural resource consumption, minimizing the risk of zoonotic disease transmission, and enabling precise control over nutritional composition and texture, offers a novel alternative source for human meat consumption. One of the major challenges to produce cultured meat in large scale is how to establish high.quality seed cells, which should have long term proliferative capacities and are able to differentiate into muscles efficienuy with simple procedures. Here, we first established an engineered porcine expanded potential stem cells (Tet-On-PAX7 EPSCs) containing Tet-On regulated PAX7 gene. Then the Tet-On-PAX7 EPSCs were induced to somite-liKe mesodermal cells. These somite-liKe mesodermal cells can be expanded over 1025-fold even after 40 passages in-vitro culture while retaining strong myogenic potential. The somite-like mesodermal cells treated with DOX for one day would differentiate into muscle stem cells (Muses), and the later were able to differentiate into muscles with an efficiency of up to 90% within just 7 days in 11-FSDeDa without Dox. Moreover, when somite-liKe mesodermal cells were seeded on patterned scaffolds, microcarrier scaffolds, or cultured in anchorage-independent suspension, they maintained high efficiency in muscle differentiation, confirming their potential to be used as seed cells for scaled cultured meat production.

Accurate monitoring of microalgae is essential for assessing marine ecological health and preventing harmful algal blooms in ocean engineering. Current in situ identification methods often suffer from limited discriminative feature extraction and inadequate adaptation to complex underwater imaging conditions. This study introduces a lightweight dual-attention neural network, termed ANMM, designed for real-time, in situ hyperspectral classification of microalgae within integrated underwater monitoring systems. The model strengthens a deepened AlexNet backbone with multi-head latent attention (MLA) and multi-head self-attention (MSA) mechanisms, which jointly enhance local feature refinement and global spectral dependency modeling. An early-stopping strategy is further incorporated to prevent overfitting and ensure robust generalization. Evaluated on a custom dataset of field-collected fluorescence spectra, the model achieves a classification accuracy of 98.91%, outperforming several state-of-the-art deep-learning counterparts. With a compact parameter size of 16.34 M and low-latency inference on edge hardware, the system demonstrates strong potential for deployment on embedded underwater sensing platforms. This work provides a practical and efficient AI-driven solution for continuous marine microalgae monitoring, supporting advances in ocean observation technology and ecological engineering.