Nat Commun. 2025 Apr 8;16(1):3340. doi: 10.1038/s41467-025-57807-5.

ABSTRACT

Longevity research aims to extend the healthspan while minimizing the duration of disability and morbidity, known as the sickspan. Most longevity interventions in model organisms extend healthspan, but it is not known whether they compress sickspan relative to the lifespan. Here, we present a theory that predicts which interventions compress relative sickspan, based on the shape of the survival curve. Interventions such as caloric restriction that extend mean lifespan while preserving the shape of the survival curve, are predicted to extend the sickspan proportionally, without compressing it. Conversely, a subset of interventions that extend lifespan and steepen the shape of the survival curve are predicted to compress the relative sickspan. We explain this based on the saturating-removal mathematical model of aging, and present evidence from longitudinal health data in mice, Caenorhabditis elegans and Drosophila melanogaster. We apply this theory to identify potential interventions for compressing the sickspan in mice, and to combinations of longevity interventions. This approach offers potential strategies for compressing morbidity and extending healthspan.

PMID:40199852 | DOI:10.1038/s41467-025-57807-5

Bioinformatics. 2025 Apr 8:btaf154. doi: 10.1093/bioinformatics/btaf154. Online ahead of print.

ABSTRACT

MOTIVATION: Single-cell time series data often exhibit significant variability within an isogenic cell population. When modeling intracellular processes, it is therefore more appropriate to infer parameter distributions that reflect this variability, rather than fitting the population average to obtain a single point estimate. The Global Two-Stage (GTS) approach for nonlinear mixed-effects (NLME) models is a simple and modular method commonly used for this purpose. However, this method is computationally intensive due to its repeated use of non-convex optimization and numerical integration of the underlying system.

RESULTS: We propose the Gradient Matching GTS (GMGTS) method as an efficient alternative to GTS. Gradient matching offers an integration-free approach to parameter estimation that is particularly powerful for systems that are linear in the unknown parameters, such as biochemical networks modeled by mass action kinetics. By incorporating gradient matching into the GTS framework, we expand its capabilities through uncertainty propagation calculations and an iterative estimation scheme for partially observed systems. Comparisons between GMGTS and GTS across various inference setups show that our method significantly reduces computational demands, facilitating the application of complex NLME models in systems biology.

AVAILABILITY AND IMPLEMENTATION: A Matlab implementation of GMGTS is provided at https://github.com/yulanvanoppen/GMGTS (DOI: http://doi.org/10.5281/zenodo.14884457).

SUPPLEMENTARY INFORMATION: Supplemental Information is available online and contains Tables S1-S4, Figures S1-S21, methodology, mathematical derivations, and software implementation details.

PMID:40199819 | DOI:10.1093/bioinformatics/btaf154

Thromb Haemost. 2025 Apr 8. doi: 10.1055/a-2561-2362. Online ahead of print.

NO ABSTRACT

PMID:40199492 | DOI:10.1055/a-2561-2362

Cancer Discov. 2025 Apr 8:OF1-OF26. doi: 10.1158/2159-8290.CD-24-0943. Online ahead of print.

ABSTRACT

This study reveals a broad convergence in genetic adaptation to hypoxia between natural populations and tumors, suggesting that insights from natural populations could enhance our understanding of cancer biology and identify novel therapeutic targets.

PMID:40199338 | DOI:10.1158/2159-8290.CD-24-0943

Curr Biol. 2025 Apr 4:S0960-9822(25)00345-8. doi: 10.1016/j.cub.2025.03.033. Online ahead of print.

ABSTRACT

Seed formation is essential for plant propagation and food production. We present a novel mechanism for the regulation of seed size by a newly identified "gate" at the chalazal end of the ovule regulating nutrient transport into the developing seed. This gate is blocked by callose deposition in unfertilized mature ovules (closed state), but the callose is removed after central cell fertilization, allowing nutrient transport into the seed (open state). However, if fertilization fails, callose deposition persists, preventing transportation of nutrients from the funiculus. A mutant in an ovule-expressed β-1,3-glucanase gene (AtBG_ppap) showed incomplete callose degradation after fertilization and produced smaller seeds, apparently due to its partially closed state. By contrast, an AtBG_ppap overexpression line produced larger seeds due to continuous callose degradation, fully opening the gate for nutrient transport into the seed. The mechanism was also identified in rice, indicating that it potentially could be applied widely to angiosperms to increase seed size.

PMID:40199323 | DOI:10.1016/j.cub.2025.03.033

Cell. 2025 Mar 25:S0092-8674(25)00192-8. doi: 10.1016/j.cell.2025.02.007. Online ahead of print.

ABSTRACT

Microglia, essential in the central nervous system (CNS), were historically considered absent from the peripheral nervous system (PNS). Here, we show a PNS-resident macrophage population that shares transcriptomic and epigenetic profiles as well as an ontogenetic trajectory with CNS microglia. This population (termed PNS microglia-like cells) enwraps the neuronal soma inside the satellite glial cell envelope, preferentially associates with larger neurons during PNS development, and is required for neuronal functions by regulating soma enlargement and axon growth. A phylogenetic survey of 24 vertebrates revealed an early origin of PNS microglia-like cells, whose presence is correlated with neuronal soma size (and body size) rather than evolutionary distance. Consistent with their requirement for soma enlargement, PNS microglia-like cells are maintained in vertebrates with large peripheral neuronal soma but absent when neurons evolve to have smaller soma. Our study thus reveals a PNS counterpart of CNS microglia that regulates neuronal soma size during both evolution and ontogeny.

PMID:40199320 | DOI:10.1016/j.cell.2025.02.007

Curr Biol. 2025 Apr 7;35(7):R251-R253. doi: 10.1016/j.cub.2025.02.057.

ABSTRACT

Germ granules are specialized RNA-protein condensates that drive germ cell development. A new study reveals that germ granules promote Drosophila germ cell formation by altering membrane mechanics through PIP2 and actin.

PMID:40199247 | DOI:10.1016/j.cub.2025.02.057

Biochem Biophys Res Commun. 2025 Apr 1;762:151745. doi: 10.1016/j.bbrc.2025.151745. Online ahead of print.

ABSTRACT

Reactive oxygen species play a crucial role in cellular processes, but their effects on protein structure and function in vivo remain challenging to study. Here, we present an approach using synchrotron-based X-ray footprinting methods to probe protein structure, via quantitative LC-coupled mass spectrometry of methionine oxidation (MSOx) in live E. coli. A label-free proteomic analysis identified 2104 proteins from E. coli, with 465 proteins exhibiting MSOx modifications distributed across multiple cellular compartments. Changes in MSOx modification with increasing X-ray dose revealed a correlation between rates of modification and solvent-accessible surface area in vivo for selected proteins responsive to exposure, providing a direct probe of protein structure and its conformational plasticity in the cell. The approach developed here offers a unique in-cell quantitative readout of methionine oxidation and solvent accessibility through radiolytic hydroxyl radical labeling. With this method, the landscape of methionine oxidation in E. coli can be mapped, providing insights into protein behavior under oxidative stress. It represents a first step in developing radiolysis and E. coli as platforms for in vivo protein structure assessment. The potential applications in drug discovery, protein engineering, and systems biology of protein conformations are considerable.

PMID:40199130 | DOI:10.1016/j.bbrc.2025.151745

ACS Synth Biol. 2025 Apr 8. doi: 10.1021/acssynbio.4c00704. Online ahead of print.

ABSTRACT

Natural biological systems use feedback regulation to effectively respond and adapt to their changing environment. Even though in engineered systems we understand how accurate feedback can be depending on the electronic or mechanical parts that it is implemented with, we largely lack a similar theoretical framework to study feedback regulation in biological systems. Specifically, it is not fully understood or quantified how accurate or robust the implementation of biological feedback actually is. In this paper, we study the sensitivity of biological feedback to variations in biochemical parameters using five example circuits: positive autoregulation, negative autoregulation, double-positive feedback, positive-negative feedback, and double-negative feedback (the toggle switch). We find that some of these examples of biological feedback are subjected to fundamental performance trade-offs, and we propose multi-objective optimization as a framework to study their properties. The impact of this work is to improve robust circuit design for synthetic biology and to improve our understanding of feedback for systems biology.

PMID:40198741 | DOI:10.1021/acssynbio.4c00704

PLoS Pathog. 2025 Apr 8;21(4):e1013053. doi: 10.1371/journal.ppat.1013053. Online ahead of print.

ABSTRACT

The tethering complexes HOPS/CORVET are central for vesicular fusion through the eukaryotic endolysosomal system, but the functions of these complexes in the intracellular development of malaria parasites are still unknown. Here we show that the HOPS/CORVET core subunits are critical for the intracellular proliferation of the malaria parasite Plasmodium falciparum. We demonstrate that HOPS/CORVET are required for parasite endocytosis and host cell cytosol uptake, as early functional depletion of the complex led to developmental arrest and accumulation of endosomes that failed to fuse to the digestive vacuole membrane. Late depletion of the core HOPS/CORVET subunits led to a severe defect in merozoite invasion as a result of the mistargeting of proteins destined to the apical secretory organelles, the rhoptries and micronemes. Ultrastructure-expansion microscopy revealed a reduced rhoptry volume and the accumulation of numerous vesicles in HOPS/CORVET deficient schizonts, further supporting a role of HOPS/CORVET in post-Golgi protein cargo trafficking to the invasion related organelles. Hence, malaria parasites have repurposed HOPS/CORVET to perform dual functions across the intraerythrocytic cycle, consistent with a canonical endocytic pathway for delivery of host cell material to the digestive vacuole in trophozoite stages and a parasite specific role in trafficking of protein cargo to the apical organelles required for invasion in schizont stages.

PMID:40198740 | DOI:10.1371/journal.ppat.1013053

PLoS Comput Biol. 2025 Apr 8;21(4):e1012919. doi: 10.1371/journal.pcbi.1012919. eCollection 2025 Apr.

ABSTRACT

Stochastic Simulation Algorithm (SSA) is crucial for modeling biochemical reactions and gene regulatory networks. Traditional SSA is characterized by Markovian property and cannot naturally model systems with time delays. Several algorithms have already been designed to handle delayed reactions, yet few easy-to-use implementations exist. To address these challenges, we have developed DelaySSA, an R package that implements currently available algorithms for SSA with or without delays. Meanwhile, we also provided Matlab and Python versions to support wider applications. We demonstrated its accuracy and validity by simulating two classical models: the Bursty model and Refractory model. We then tested its capability to simulate the RNA Velocity model, where it successfully reproduced both the up- and down-regulation stages in the phase portrait. Finally, we extended its application to simulate a gene regulatory network of lung cancer adeno-to-squamous transition (AST) and qualitatively analyzed its bistability behavior by approximating the Waddington's landscape. Modeling the therapeutic intervention of a SOX2 degrader as a delayed degradation reaction, AST is effectively blocked and reprogrammed back to the adenocarcinoma state, providing a useful clue for targeting drug-resistant AST in the future. Taken together, DelaySSA is a powerful and easy-to-use software suite, facilitating accurate modeling of various kinds of biological systems and broadening the scope of stochastic simulations in systems biology.

PMID:40198732 | DOI:10.1371/journal.pcbi.1012919

Alzheimers Res Ther. 2025 Apr 8;17(1):77. doi: 10.1186/s13195-025-01728-4.

ABSTRACT

In this article, we have carefully read the author's comments on our published article regarding the post-marketing safety concerns of lecanemab based on the Food and Drug Administration Adverse Event Reporting System (FAERS) database. Pharmacovigilance studies based on the disproportionality analysis through the case/non-case design are common, and the details of this method deserve attention. We acknowledged the author's perspectives on the term "signal of disproportionate reporting (SDR)", and make some explanations on the SDR results for pancreatic carcinoma and the deduplication methods.

PMID:40200347 | DOI:10.1186/s13195-025-01728-4

BMC Endocr Disord. 2025 Apr 8;25(1):94. doi: 10.1186/s12902-025-01914-3.

ABSTRACT

BACKGROUND AND AIM: Hypothyroidism (HoT) treatment involves lifelong thyroxine replacement therapy and regular monitoring. The objective of this study was to assess the impact of clinical pharmacist (CP) intervention in managing drug-related problems (DRPs) on outcomes among patients with HoT receiving levothyroxine (LT4) therapy.

METHOD: A randomized controlled trial involved patients with HoT attending a university hospital's endocrinology and metabolism outpatient clinic from March 2022 to September 2022. Participants were randomly assigned to control (CG) and intervention groups (IG). CP identified and classified DRPs based on Pharmaceutical Care Network Europe (PCNE) v9.1 criteria. The validated version of the Morisky-Green-Levine (MGL) 4-question scale was used to measure adherence. All patients included in the study were assessed during their first visit and again two months later at their second visit.

RESULTS: 43 patients were assigned to the CG (n = 25) and IG (n = 18). Diabetes (21.6 vs. 20.5%) and hypertension (16.2% vs. 11.7%) were the most prevalent comorbidities in both the CG and IG, respectively. A total of 118 DRPs belonging to both groups were detected. In the IG group, the total number of DRPs significantly decreased from 66 to 24, and the total potential drug-drug interactions (pDDIs) decreased from 21 to 0 between the first and second visits (p < 0.001). CG and IG patients had no difference in adherence levels at the first and second visits (p > 0.05). A statistically significant increase in adherence to the time of taking the medication was observed between the first and second visits in IG (55.5% vs. 94.4%, p = 0.008).

CONCLUSION: This study highlights the frequent occurrence of DRPs and LT4 therapy adherence problems in patients with HoT. The findings suggest that the intervention of CPs, by increasing adherence to LT4 therapy and decreasing DRPs, could significantly contribute to improving patients' treatment outcomes.

TRIAL REGISTRATION: This study protocol has been retrospectively registered at ClinicalTrials.gov (NCT06408909) at 06/05/2024.

PMID:40200273 | DOI:10.1186/s12902-025-01914-3

BMC Psychiatry. 2025 Apr 8;25(1):342. doi: 10.1186/s12888-025-06646-1.

ABSTRACT

BACKGROUND: This clinical pilot initiative, led by the Sichuan Provincial Government in western China, aims to develop an innovative model for "integrated hospital-community management and treatment of severe mental disorders" to enhance the management of patients with such conditions. This single-arm, open-label, prospective, multicenter, interventional study was designed to evaluate the efficacy and safety of paliperidone palmitate 1-month formulation (PP1M) in patients with schizophrenia.

METHODS: The study enrolled patients with schizophrenia aged 18-55 from all 21 prefectural-level municipalities in Sichuan Province, southwestern China, who received PP1M at doses of 75, 100, or 150 mg. Efficacy and social functioning were assessed using the PANSS and SDSS scales at baseline and after the 3rd, 6th, and 9th injections. The SF-12 scale was administered at baseline and at 6 months, while satisfaction (MSQ) was measured at 1, 3, 6, and 9 months. The TESS scale was performed at the 1st, 3rd, 6th, and 9th doses, with adverse events (AEs) monitored after the 1st, 2nd, 3rd, and 9th doses to evaluate safety.

RESULTS: A total of 2268 patients were enrolled, with 1491 completing the 9th injection. The numbers of patients completing the PANSS, SDSS, SF-12, MSQ, and TESS scales were 1151, 1158, 1043, 827, and 1158, respectively. Following multiple PP1M doses, total PANSS scores, positive symptom subscale scores, negative symptom subscale scores, general psychopathological subscale scores, and SDSS scores significantly decreased compared to baseline (P < 0.05), while SF-12 and MSQ scores significantly increased (P < 0.05). The PANSS total response rates were 64.90% (747/1151), 74.63% (859/1151), and 78.71% (906/1151) after the 3rd, 6th, and 9th injections, respectively. TESS scores significantly decreased compared to the first injection (P < 0.05), reflecting a lower incidence of AEs. Common adverse reactions, including tremors, muscle tension, dizziness, and fatigue, were mild to moderate in severity.

CONCLUSIONS: These findings indicate that PP1M significantly alleviates psychotic symptoms in patients with schizophrenia, accelerates social function recovery, and improves quality of life and patient satisfaction. Furthermore, PP1M demonstrates improved tolerability, fewer adverse reactions, and no new drug-related safety concerns.

PMID:40200201 | DOI:10.1186/s12888-025-06646-1

Microbiol Spectr. 2025 Apr 8:e0233224. doi: 10.1128/spectrum.02332-24. Online ahead of print.

ABSTRACT

Tuberculosis remains the deadliest infectious disease of the 21st century. New antimicrobials are needed to improve treatment outcomes and enable therapy shortening. Drug repurposing is an alternative to the traditional drug discovery process. The avermectins are a family of macrocyclic lactones with anthelmintic activity active against Mycobacterium tuberculosis. However, their mode of action in mycobacteria remains unknown. In this study, we employed traditional mutant isolation approaches using Mycobacterium smegmatis, a non-pathogenic M. tuberculosis surrogate. We were only able to isolate mutants with decreased susceptibility to selamectin using the ∆nucS mutator M. smegmatis strain. This phenotype was caused by mutations in mps1 and mmpL11. Two of these mutants were used for a second experiment in which high-level selamectin-resistant mutants were isolated; however, specific mutations driving the phenotypic change to high-level resistance could not be identified. The susceptibility to selamectin in these mutants was restored to the basal level by subinhibitory concentrations of ethambutol. The selection of ethambutol resistance in a high-level selamectin-resistant mutant also resulted in multiple colonies becoming susceptible to selamectin again. These colonies carried mutations in embB, suggesting that the integrity of the cell envelope is a prerequisite for selamectin resistance. The absence of increased susceptibility to selamectin in an embB deletion strain demonstrated that the target of selamectin is not cytosolic. Our data show that the concurrence of specific multiple mutations and complete integrity of the mycobacterial envelope are necessary for selamectin resistance. Our studies provide first-time insights into the antimycobacterial mode of action of the antiparasitic avermectins.IMPORTANCETuberculosis is the deadliest infectious disease of the 21st century. New antibiotics are needed to improve treatment. However, developing new drugs is costly and lengthy. Drug repurposing is an alternative to the traditional drug discovery process. The avermectins are a family of drugs used to treat parasitic infections that are active against Mycobacterium tuberculosis, the bacterium that causes tuberculosis. However, their mode of action in mycobacteria remains unknown. Understanding how avermectins kill mycobacteria can facilitate its development as an anti-mycobacterial drug, including against M. tuberculosis.In this study, we used Mycobacterium smegmatis, a non-pathogenic M. tuberculosis surrogate model to understand the molecular mechanisms of how selamectin (a drug of the avermectin family selected for this study as a model) acts against mycobacteria. Our data show that the generation of resistance to selamectin is unlikely and that complete integrity of the mycobacterial envelope is necessary for selamectin resistance, providing first-time insights into the antimycobacterial mode of action of the avermectins.

PMID:40197087 | DOI:10.1128/spectrum.02332-24

mBio. 2025 Apr 8:e0064625. doi: 10.1128/mbio.00646-25. Online ahead of print.

ABSTRACT

Pseudomonas aeruginosa is a Gram-negative opportunistic pathogen that poses a significant public health threat, particularly in healthcare settings. A key determinant of P. aeruginosa virulence is the regulated synthesis and release of extracellular products, which is controlled by a cell density-dependent signaling system known as quorum sensing (QS). P. aeruginosa uses a complex QS network, including two systems that rely on diffusible N-acylhomoserine lactone (AHL) signal molecules. The LuxR-type receptor RhlR is unique in that it requires not only its cognate AHL but also the accessory protein PqsE to maximally bind to promoter DNA and initiate transcription. Our group previously demonstrated that PqsE physically interacts with RhlR, enhancing its affinity for target promoters across the P. aeruginosa genome. Although LuxR-type receptors are widespread in Gram-negative bacteria and important for pathogenesis, PqsE orthologs are restricted to Pseudomonas and Burkholderia species. This study explored the conservation of PqsE and examined PqsE ortholog structure-function across different species. Our results show that PqsE in Pseudomonas retains their functional interactions with RhlR homologs, unlike PqsE orthologs in Burkholderia spp., which do not interact with their respective LuxR-type receptors. Additionally, we assessed the AHL preferences of different receptors and hypothesized that the PqsE-RhlR interaction evolved to stabilize the inherently unstable RhlR, preventing its degradation. Indeed, we observe higher levels of RhlR protein turnover in a strain lacking pqsE compared to a wild-type strain of PA14, which can be partially rescued in a strain of P. aeruginosa lacking the Lon protease.

IMPORTANCE: Pseudomonas aeruginosa, a major pathogen for patients with cystic fibrosis and a primary constituent of healthcare-associated infections, relies on a complex quorum-sensing (QS) network to coordinate virulence factor production. Central to this system is the interaction between two proteins, PqsE and RhlR, which drive gene expression essential for pathogenesis. Our study investigates the conservation of the PqsE-RhlR interaction across related bacterial species, revealing that PqsE in Pseudomonas can enhance RhlR activity, while orthologs in Burkholderia lack this capacity. These findings offer new insights into the specificity and evolution of QS mechanisms, highlighting the PqsE-RhlR interaction as a potentially selective target for treating P. aeruginosa infections.

PMID:40197035 | DOI:10.1128/mbio.00646-25

STAR Protoc. 2025 Apr 7;6(2):103738. doi: 10.1016/j.xpro.2025.103738. Online ahead of print.

ABSTRACT

Functional impact of noncoding variants can be predicted using computational approaches. Although predictive scores can be insightful, implementing the scores for a custom variant set and associating scores with complex traits require multiple phases of analysis. Here, we present a protocol for prioritizing variants by generating deep-learning-predicted functional scores and relating them with brain traits. We describe steps for score prediction, statistical comparison, phenotype correlation, and functional enrichment analysis. This protocol can be generalized to different models and phenotypes. For complete details on the use and execution of this protocol, please refer to Mondragon-Estrada et al.1.

PMID:40198216 | DOI:10.1016/j.xpro.2025.103738

J Gerontol B Psychol Sci Soc Sci. 2025 Apr 8:gbaf065. doi: 10.1093/geronb/gbaf065. Online ahead of print.

ABSTRACT

OBJECTIVE: Alzheimer's disease and related dementias significantly impact older adults' quality of life. The clock-drawing test (CDT) is a widely used dementia screening tool due to its ease of administration and effectiveness. However, manual CDT-coding in large-scale studies can be time-intensive and prone to coding errors and is typically limited to ordinal responses. In this study, we developed a continuous CDT score using a deep learning neural network (DLNN) and evaluated its ability to classify participants as having dementia or not.

METHODS: Using a nationally representative sample of older adults from the National Health and Aging Trends Study (NHATS), we trained deep learning models on CDT images to generate both ordinal and continuous scores. Using a modified NHATS dementia classification algorithm as a benchmark, we computed the Area Under the Receiver Operating Characteristic Curve for each scoring approach. Thresholds were determined by balancing sensitivity and specificity, and demographic-specific thresholds were compared to a uniform threshold for classification accuracy.

RESULTS: Continuous CDT scores provided more granular thresholds than ordinal scores for dementia classification, which vary by demographic characteristics. Lower thresholds were identified for Black individuals, those with lower education, and those ages 90 or older. Compared to ordinal scores, continuous scores also allowed for a more balanced sensitivity and specificity.

DISCUSSION: This study demonstrates the potential of continuous CDT generated by DLNN to enhance dementia classification. By identifying demographic-specific thresholds, it offers a more inclusive and adaptive approach, which could lead to improved guidelines for using CDT in dementia screening.

PMID:40197801 | DOI:10.1093/geronb/gbaf065

J Chem Phys. 2025 Apr 14;162(14):144101. doi: 10.1063/5.0257558.

ABSTRACT

We investigate the locality of magnetic response in polycyclic aromatic molecules using a novel deep-learning approach. Our method employs graph neural networks (GNNs) with a graph-of-rings representation to predict nucleus independent chemical shifts (NICS) in the space around the molecule. We train a series of models, each time reducing the size of the largest molecules used in training. The accuracy of prediction remains high (MAE < 0.5 ppm), even when training the model only on molecules with up to four rings, thus providing strong evidence for the locality of magnetic response. To overcome the known problem of generalization of GNNs, we implement a k-hop expansion strategy and succeed in achieving accurate predictions for molecules with up to 15 rings (almost 4 times the size of the largest training example). Our findings have implications for understanding the magnetic response in complex molecules and demonstrate a promising approach to overcoming GNN scalability limitations. Furthermore, the trained models enable rapid characterization, without the need for more expensive DFT calculations.

PMID:40197568 | DOI:10.1063/5.0257558

BMC Genomics. 2025 Apr 7;26(1):350. doi: 10.1186/s12864-025-11511-2.

ABSTRACT

BACKGROUND: Clustering scRNA-seq data plays a vital role in scRNA-seq data analysis and downstream analyses. Many computational methods have been proposed and achieved remarkable results. However, there are several limitations of these methods. First, they do not fully exploit cellular features. Second, they are developed based on gene expression information and lack of flexibility in integrating intercellular relationships. Finally, the performance of these methods is affected by dropout event.

RESULTS: We propose a novel deep learning (DL) model based on attention autoencoder and zero-inflated (ZI) layer, namely scAMZI, to cluster scRNA-seq data. scAMZI is mainly composed of SimAM (a Simple, parameter-free Attention Module), autoencoder, ZINB (Zero-Inflated Negative Binomial) model and ZI layer. Based on ZINB model, we introduce autoencoder and SimAM to reduce dimensionality of data and learn feature representations of cells and relationships between cells. Meanwhile, ZI layer is used to handle zero values in the data. We compare the performance of scAMZI with nine methods (three shallow learning algorithms and six state-of-the-art DL-based methods) on fourteen benchmark scRNA-seq datasets of various sizes (from hundreds to tens of thousands of cells) with known cell types. Experimental results demonstrate that scAMZI outperforms competing methods.

CONCLUSIONS: scAMZI outperforms competing methods and can facilitate downstream analyses such as cell annotation, marker gene discovery, and cell trajectory inference. The package of scAMZI is made freely available at https://doi.org/10.5281/zenodo.13131559 .

PMID:40197174 | DOI:10.1186/s12864-025-11511-2