Nat Commun. 2025 Feb 3;16(1):1284. doi: 10.1038/s41467-025-56475-9.

ABSTRACT

Cells are regulated at multiple levels, from regulations of individual genes to interactions across multiple genes. Some recent neural network models can connect molecular changes to cellular phenotypes, but their design lacks modeling of regulatory mechanisms, limiting the decoding of regulations behind key cellular events, such as cell state transitions. Here, we present regX, a deep neural network incorporating both gene-level regulation and gene-gene interaction mechanisms, which enables prioritizing potential driver regulators of cell state transitions and providing mechanistic interpretations. Applied to single-cell multi-omics data on type 2 diabetes and hair follicle development, regX reliably prioritizes key transcription factors and candidate cis-regulatory elements that drive cell state transitions. Some regulators reveal potential new therapeutic targets, drug repurposing possibilities, and putative causal single nucleotide polymorphisms. This method to analyze single-cell multi-omics data demonstrates how the interpretable design of neural networks can better decode biological systems.

PMID:39900922 | DOI:10.1038/s41467-025-56475-9

Strahlenther Onkol. 2025 Feb 3. doi: 10.1007/s00066-024-02364-x. Online ahead of print.

ABSTRACT

BACKGROUND: Adjuvant radiotherapy is essential for reducing local recurrence and improving survival in breast cancer patients, but it carries a risk of ischemic cardiac toxicity, which increases with heart exposure. The isocentric lateral decubitus position, where the breast rests flat on a support, reduces heart exposure and leads to delivery of a more uniform dose. This position is particularly beneficial for patients with unique anatomies, such as those with pectus excavatum or larger breast sizes. While artificial intelligence (AI) algorithms for autocontouring have shown promise, they have not been tailored to this specific position. This study aimed to develop and evaluate a neural network-based autocontouring algorithm for patients treated in the isocentric lateral decubitus position.

MATERIALS AND METHODS: In this single-center study, 1189 breast cancer patients treated after breast-conserving surgery were included. Their simulation CT scans (1209 scans) were used to train and validate a neural network-based autocontouring algorithm (nnU-Net). Of these, 1087 scans were used for training, and 122 scans were reserved for validation. The algorithm's performance was assessed using the Dice similarity coefficient (DSC) to compare the automatically delineated volumes with manual contours. A clinical evaluation of the algorithm was performed on 30 additional patients, with contours rated by two expert radiation oncologists.

RESULTS: The neural network-based algorithm achieved a segmentation time of approximately 4 min, compared to 20 min for manual segmentation. The DSC values for the validation cohort were 0.88 for the treated breast, 0.90 for the heart, 0.98 for the right lung, and 0.97 for the left lung. In the clinical evaluation, 90% of the automatically contoured breast volumes were rated as acceptable without corrections, while the remaining 10% required minor adjustments. All lung contours were accepted without corrections, and heart contours were rated as acceptable in 93.3% of cases, with minor corrections needed in 6.6% of cases.

CONCLUSION: This neural network-based autocontouring algorithm offers a practical, time-saving solution for breast cancer radiotherapy planning in the isocentric lateral decubitus position. Its strong geometric performance, clinical acceptability, and significant time efficiency make it a valuable tool for modern radiotherapy practices, particularly in high-volume centers.

PMID:39900818 | DOI:10.1007/s00066-024-02364-x

Orthopadie (Heidelb). 2025 Feb 3. doi: 10.1007/s00132-025-04619-6. Online ahead of print.

ABSTRACT

BACKGROUND: Artificial intelligence is very likely to be a pioneering technology in arthroplasty, with a wide range of pre-, intra- and post-operative applications. The opportunities for patients, doctors and healthcare policy are considerable, especially in the context of optimized and individualized patient care.

DATA AVAILABILITY: Despite these diverse possibilities, there are currently only a few AI applications in routine clinical practice, mainly due to the limited availability of analyzable health data. AI systems are only as good as the data they are trained with. If the data is insufficient, incomplete or biased, the AI may draw false conclusions. The current results of such AI applications in arthroplasty must, therefore, be viewed critically, especially as previous data bases were not designed a priori for AI applications.

PROSPECTS: The successful integration of AI, therefore, requires a targeted focus on the development of a specific data structure. In order to exploit the full potential of AI, comprehensive clinical data volumes are required, which can only be realized through a multicentric approach. In this context, ethical and data protection issues remain a further question, and not only in orthopaedics. Cooperative efforts at national and international levels are, therefore, essential in order to research and develop new AI applications.

PMID:39900780 | DOI:10.1007/s00132-025-04619-6

Sci Rep. 2025 Feb 3;15(1):4120. doi: 10.1038/s41598-024-76561-0.

ABSTRACT

Idiopathic pulmonary fibrosis (IPF) is a chronic lung disease characterized by excessive scarring and fibrosis due to the abnormal accumulation of extracellular matrix components, primarily collagen. This study aims to design and solve an optimal control problem to regulate M2 macrophage activity in IPF, thereby preventing fibrosis formation by controlling the anti-M1 agent. The research models the diffusion of M2 macrophages in inflamed tissue using a novel dynamical system with partial differential equation (PDE) constraints. The control problem is formulated to minimize fibrosis by regulating an anti-M1 agent. The study employs a two-step process of discretization followed by optimization, utilizing the Galerkin spectral method to transform the M2 diffusion PDE into an algebraic system of ordinary differential equations (ODEs). The optimal control problem is then solved using Pontryagin/s minimum principle, canonical Hamiltonian equations, and extended Riccati differential equations. The numerical simulations indicate that without control, M2 macrophage levels increase and stabilize, contributing to fibrosis. In contrast, the optimal control strategy effectively reduces M2 macrophages, preventing fibrosis formation within 120 days. The results highlight the potential of the proposed optimal control approach in modulating tissue repair processes and mitigating the progression of IPF. This study underscores the significance of targeting M2 macrophages and employing mathematical methods to develop innovative therapies for lung fibrosis.

PMID:39900943 | DOI:10.1038/s41598-024-76561-0

Nat Commun. 2025 Feb 3;16(1):1303. doi: 10.1038/s41467-025-56501-w.

ABSTRACT

Irreversible damage of the lung epithelium in idiopathic pulmonary fibrosis (IPF) patients causes high mortality worldwide, with no lung repair approaches available currently. Here we show that in murine and monkey models, the KRT5+ P63+ progenitor cells in airway basal layer can enter the alveolar area post fibrotic injury. Aided with an automated culture system, we clone and characterize airway basal progenitor cells from 44 donors with various lung conditions. Transplantation of human progenitor cells into the mouse lung efficiently re-epithelializes the injured alveolar area, forms new respiratory tract and saccule-like structures, which ameliorates fibrotic lesions and improves survival of mice. Mechanistically, the engrafted human progenitor cells do not function by differentiating into mature alveolar cells in mouse lung; instead, they differentiate into saccular cells expressing multiple tight junction proteins such as CLDN4, which help the lung to re-establish epithelial barriers. Furthermore, by cloning P63+ airway basal progenitors from larger mammals and birds, we construct multiple lung-chimerism animals and uncover the evolutionarily conserved roles of these progenitor cells in lung repair. Overall, our data highlight the fate of airway basal progenitor cells in fibrotic lung and provide a potential therapeutic strategy for pulmonary diseases that lack inherent recovery mechanisms.

PMID:39900892 | DOI:10.1038/s41467-025-56501-w

BMC Gastroenterol. 2025 Feb 3;25(1):50. doi: 10.1186/s12876-025-03629-0.

ABSTRACT

BACKGROUND: Ulcerative colitis (UC) is a type of inflammatory bowel disease, and current treatments often fall short, necessitating new therapeutic options. Baicalin shows therapeutic promise in UC animal models, but a systematic review is needed.

METHODS: A systematic search was conducted across databases including PubMed, EBSCO, Web of Science, and Science Direct, up to March 2024, identifying randomized controlled trials (RCTs) examining baicalin's impact on UC in animal models. Seventeen studies were selected through manual screening. Meta-analyses and subgroup analyses utilized Rev Man 5.3 and Stata 15.0 software to assess symptom improvement.

RESULTS: From 1304 citations, 17 were analyzed. Baicalin significantly modulated various biomarkers: HCS (SMD = -3.91), DAI (MD = -2.75), spleen index (MD = -12.76), MDA (SMD = -3.88), IL-6 (SMD = -10.59), IL-1β (SMD = -3.98), TNF-α (SMD = -8.05), NF-κB (SMD = -5.46), TLR4 (MD = -0.38), RORγ (MD = -0.89), MCP-1 (MD = -153.25), MPO (SMD = -7.34), Caspase-9 (MD = -0.93), Caspase-3 (MD = -0.45), FasL (MD = -1.20)) and enhanced BWC (MD = 0.06), CL (MD = 1.39), ZO-1 (MD = 0.44), SOD (SMD = 3.04), IL-10 mRNA (MD = 3.14), and FOXP3 (MD = 0.45) levels. Baicalin's actions may involve the PI3K/AKT, TLR4/NF-κB, IKK/IKB, Bcl-2/Bax, Th17/Treg, and TLRs/MyD88 pathways. Optimal therapeutic outcomes were predicted at dosages of 60-150 mg/kg over 10-14 weeks.

CONCLUSION: Baicalin demonstrates a multifaceted therapeutic potential in UC, attributed to its anti-inflammatory, antioxidant, anti-apoptotic, and intestinal barrier repair properties. While higher doses and longer treatments appear beneficial, further research, particularly human clinical trials, is necessary to verify its effectiveness and safety in people.

PMID:39901089 | DOI:10.1186/s12876-025-03629-0

Nat Microbiol. 2025 Feb;10(2):527-540. doi: 10.1038/s41564-024-01914-4. Epub 2025 Feb 3.

ABSTRACT

The local arrangement of microbes can profoundly impact community assembly, function and stability. However, our understanding of the spatial organization of the human gut microbiome at the micron scale is limited. Here we describe a high-throughput and streamlined method called Split-And-pool Metagenomic Plot-sampling sequencing (SAMPL-seq) to capture spatial co-localization in a complex microbial consortium. The method obtains microbial composition of micron-scale subcommunities through split-and-pool barcoding. SAMPL-seq analysis of the healthy human gut microbiome identified bacterial taxa pairs that consistently co-occurred both over time and across multiple individuals. These co-localized microbes organize into spatially distinct groups or 'spatial hubs' dominated by Bacteroidaceae, Ruminococcaceae and Lachnospiraceae families. Using inulin as a dietary perturbation, we observed reversible spatial rearrangement of the gut microbiome where specific taxa form new local partnerships. Spatial metagenomics using SAMPL-seq can unlock insights into microbiomes at the micron scale.

PMID:39901058 | DOI:10.1038/s41564-024-01914-4

Nat Biotechnol. 2025 Feb 3. doi: 10.1038/s41587-025-02576-1. Online ahead of print.

NO ABSTRACT

PMID:39901026 | DOI:10.1038/s41587-025-02576-1

Nat Genet. 2025 Feb 3. doi: 10.1038/s41588-024-02069-y. Online ahead of print.

ABSTRACT

A pan-transcriptome describes the transcriptional and post-transcriptional consequences of genome diversity from multiple individuals within a species. We developed a barley pan-transcriptome using 20 inbred genotypes representing domesticated barley diversity by generating and analyzing short- and long-read RNA-sequencing datasets from multiple tissues. To overcome single reference bias in transcript quantification, we constructed genotype-specific reference transcript datasets (RTDs) and integrated these into a linear pan-genome framework to create a pan-RTD, allowing transcript categorization as core, shell or cloud. Focusing on the core (expressed in all genotypes), we observed significant transcript abundance variation among tissues and between genotypes driven partly by RNA processing, gene copy number, structural rearrangements and conservation of promotor motifs. Network analyses revealed conserved co-expression module::tissue correlations and frequent functional diversification. To complement the pan-transcriptome, we constructed a comprehensive cultivar (cv.) Morex gene-expression atlas and illustrate how these combined datasets can be used to guide biological inquiry.

PMID:39901014 | DOI:10.1038/s41588-024-02069-y

Nat Commun. 2025 Feb 3;16(1):1277. doi: 10.1038/s41467-025-56489-3.

ABSTRACT

Eukaryotic translation elongation factor 1A (eEF1A) is a highly abundant, multi-domain GTPase. Post-translational steps essential for eEF1A biogenesis are carried out by bespoke chaperones but co-translational mechanisms tailored to eEF1A folding remain unexplored. Here, we use AlphaPulldown to identify Ypl225w (also known as Chp1, Chaperone 1 for eEF1A) as a conserved yeast protein predicted to stabilize the N-terminal, GTP-binding (G) domain of eEF1A against its misfolding propensity, as predicted by computational simulations and validated by microscopy analysis of ypl225wΔ cells. Proteomics and biochemical reconstitution reveal that Ypl225w functions as a co-translational chaperone by forming dual interactions with the eEF1A G domain nascent chain and the UBA domain of ribosome-bound nascent polypeptide-associated complex (NAC). Lastly, we show that Ypl225w primes eEF1A nascent chains for binding to GTP as part of a folding mechanism tightly coupled to chaperone recycling. Our work shows that an ATP-independent chaperone can drive vectorial folding of nascent chains by co-opting G protein nucleotide binding.

PMID:39900909 | DOI:10.1038/s41467-025-56489-3

NPJ Precis Oncol. 2025 Feb 3;9(1):35. doi: 10.1038/s41698-025-00804-0.

ABSTRACT

The rapid development of deep learning has revolutionized medical image processing, including analyzing whole slide images (WSIs). Despite the demonstrated potential for characterizing gene mutations directly from WSIs in certain cancers, challenges remain due to image resolution and reliance on manual annotations for acute myeloid leukemia (AML). We, therefore, propose a deep learning model based on multiple instance learning (MIL) with ensemble techniques to predict gene mutations from AML WSIs. Our model predicts NPM1 mutations and FLT3-ITD without requiring patch-level or cell-level annotations. Using a dataset of 572 WSIs, the largest database with both WSI and genetic mutation information, our model achieved an AUC of 0.90 ± 0.08 for NPM1 and 0.80 ± 0.10 for FLT3-ITD in the testing cohort. Additionally, we found that blasts are pivotal indicators for gene mutation predictions, with their proportions varying between mutated and standard WSIs, highlighting the clinical potential of AML WSI analysis.

PMID:39900774 | DOI:10.1038/s41698-025-00804-0

Methods Mol Biol. 2025;2880:331-344. doi: 10.1007/978-1-0716-4276-4_17.

ABSTRACT

A significantly low success rate of human clinical studies has long been attributed to a capability gap, namely, an ineffective translation of the animal data to the human context. To bridge this capability gap, several correcting measures have been evaluated; using a strict guideline to select animal models for a given disease and implementing alternative models such as tissues-on-chip are some of them. Current hypothesis tells that there is a basic similarity in responding to a stress between human and those mammals that precede human in the phylogenetic tree; however, the corresponding molecular mechanisms are not exactly the same across these species. Therefore, strategic manipulations are necessary to curate those candidates from animal data that would have high translational potential. Hence, we developed an analytical tool that can screen the in vivo results, such as genomic, proteomic, epigenomic data with two primary objectives. The first objective is to identify those molecules that are sequentially conserved across the phylogenetic tree. The second objective is to find those molecules that would similarly perturb across the phylogenetic tree in responding to a stress of interest. A machine learning (ML) algorithm converges these two sets of molecules to curate the common features, which would demonstrate phylogenetic homology in their molecular makeups and characteristic similarity across the phylogenetic tree. This ML-pipeline would be most beneficial in those scenarios, such as the rare diseases or chemical-biological-radiation-nuclear (CBRN)-exposed samples, where the inventory of human samples is minimum. This strategy is surely at a risk in overlooking the human-exclusive signatures; nevertheless, this ML-approach is poised to refine the animal data to generate results of high translational potential with minimum false positive and false negative entries.

PMID:39900768 | DOI:10.1007/978-1-0716-4276-4_17

Methods Mol Biol. 2025;2880:69-99. doi: 10.1007/978-1-0716-4276-4_4.

ABSTRACT

The discovery of genes that code for a specific enzymatic activity is important in various fields of life science and provides valuable biotechnological tools. Many genes that contribute to the production of secondary metabolites and specialized metabolic pathways are still not identified. Due to the great diversity of metabolic functions found in nature and their rapid evolutionary adaptation, we need precise but high-throughput approaches for a targeted search based on minimal prior knowledge. In this chapter, we describe a transcriptomics pipeline that was used to search for candidate genes coding for a specific enzymatic activity in a nonmodel species. We generated and combined short- and long-read transcriptomic data to obtain reliable full-length transcript sequences along with information on allelic variation, isoform expression, and condition-specific expression. Based on protein domain annotations of coding sequences and transcriptomic data, we selected candidate genes for activity assays. We provide detailed instructions for analysis and quality control steps in our pipeline that can be applied to other biological questions.

PMID:39900755 | DOI:10.1007/978-1-0716-4276-4_4

Methods Mol Biol. 2025;2880:1-16. doi: 10.1007/978-1-0716-4276-4_1.

ABSTRACT

Saliva, a readily available and noninvasive biofluid, has emerged as a promising source for gene expression studies, offering a window into both local and systemic gene expression patterns. The salivary transcriptome and miRNome hold valuable information about the physiological and pathological processes occurring in the oral cavity and throughout the body.This chapter delves into the potential of saliva as a noninvasive sampling method, exploring its utility in gene expression profiling for various applications. It provides an overview of the components contributing to the salivary transcriptome and discusses the challenges associated with salivary RNA analysis. We highlight the applications of salivary gene expression studies in biomarker discovery for oral and systemic diseases.While discussing various saliva collection techniques, here we focus on the procedure for RNA extraction, including microRNA (miRNA) from the OMNIgene™ SALIVA DNA and RNA device, OMR-610 (DNA Genotek Inc., Ottawa, Ontario, Canada). Herein, we provide the detailed methodologies for RNA extraction for salivary transcriptomics and the miRNome, thus providing a resource for researchers interested in leveraging the diagnostic and prognostic potential of saliva for personalized medicine and precision health initiatives.

PMID:39900752 | DOI:10.1007/978-1-0716-4276-4_1

Sci Rep. 2025 Feb 3;15(1):4104. doi: 10.1038/s41598-025-88838-z.

ABSTRACT

This study aims to monitor and identify adverse events (AEs) associated with cetuximab, a drug used to treat various late-stage (metastatic) tumors, to improve patient safety and guide drug use. This study retrospectively analyzed the cases reported in the FDA adverse event reporting system (FAERS) related to the application of cetuximab from 2013 Q1 to 2022 Q4. Disproportionality analyses, including the reporting odds ratio (ROR), the proportional reporting ratio (PRR), the Bayesian confidence propagation neural network (BCPNN), and the empirical Bayesian geometric mean (EBGM) algorithms, were employed to quantify the signals of cetuximab-associated AEs. A total of 8364225 reports were contained in the FAERS database, of which 5186 reports of cetuximab were identified as 'primary suspected (PS)' AEs. The application of cetuximab resulted in AEs in 22 system organ classes (SOCs), which preserved 176 significant disproportionality preferred terms (PTs) through the computation of four algorithms. The main SOCs (Skin and subcutaneous tissue disorders, investigations, metabolism and nutrition disorders, and blood and lymphatic system disorders) accounted for 58.63%. Some AEs were not on the drug label: speech disorder, intervertebral discitis, glomerulonephritis rapidly progressive and disseminated intravascular coagulation. This study identified new signals of adverse drug reactions (ADRs) other than those mentioned in the specification associated with cetuximab, providing valuable insights into the relationship between ADRs and cetuximab use. The findings highlight the importance of continuous surveillance to detect and manage AEs effectively, ultimately improving patient safety during treatment with cetuximab.

PMID:39901061 | DOI:10.1038/s41598-025-88838-z

Drug Ther Bull. 2025 Feb 3:dtb-2024-000073. doi: 10.1136/dtb.2024.000073. Online ahead of print.

NO ABSTRACT

PMID:39900487 | DOI:10.1136/dtb.2024.000073

Pharm Dev Technol. 2025 Feb 3:1-45. doi: 10.1080/10837450.2025.2462946. Online ahead of print.

ABSTRACT

Cinacalcet hydrochloride (HCl), a calcium-sensing receptor agonist used to treat hyperparathyroidism, suffers from poor solubility, reducing its bioavailability. Recently, cinacalcet HCl has been probed for repurposing as antibacterial agent. This work investigates cinacalcet HCl's potential as an antibacterial agent and provides a formulation to improve the drug dissolution. Solid dispersion formulations using Poloxamer 407, with and without Soluplus®, were prepared via solvent evaporation and hot melt congealing methods. The resulting formulations were analyzed using differential scanning calorimetry, FTIR spectroscopy, X-ray powder diffraction, and dissolution studies. These formulations significantly enhanced cinacalcet HCl dissolution compared to the unprocessed form, achieving up to a 15-fold increase in Q5 (percent of cinacalcet HCl dissolved after 5 minutes). The dissolution efficiency rose from 28% for the pure drug to 94.8% and 87.8% for formulations F6 and F7, respectively. Microbiological evaluations confirmed the antibacterial effect of cinacalcet HCl, which was notably increased in the Poloxamer 407 and Soluplus® hybrid formulation (F7) with a MIC of 64-128 µg/ml. Antibiofilm activity was also observed, with qRT-PCR indicating downregulation of biofilm genes (icaA, icaD, and fnbA). This study introduces a cinacalcet HCl formulation prepared using a scalable, green approach, demonstrating significant potential for antimicrobial applications.

PMID:39899403 | DOI:10.1080/10837450.2025.2462946

Clin Transl Sci. 2025 Feb;18(2):e70108. doi: 10.1111/cts.70108.

ABSTRACT

5-hydroxytryptamine-3 (5-HT3) receptor antagonists including ondansetron, tropisetron, dolasetron, palonosetron, granisetron, and ramosetron are commonly used to prevent and treat nausea and vomiting. Most of these medications are at least partially metabolized via the highly polymorphic CYP2D6 enzyme, resulting in variations of metabolism among individuals. Current (2017) international prescribing guidelines for ondansetron/tropisetron use according to genotype provide recommendations for CYP2D6 ultrarapid metabolizers but not intermediate or poor metabolizers. However, multiple studies have been conducted since this guideline was published. This review evaluated all available evidence of an association between CYP2D6 genotype and 5-HT3 receptor antagonist outcomes, including in patients who are CYP2D6 intermediate/poor metabolizers and pediatric-specific studies. In this review, we confirm that CYP2D6 genotype impacts ondansetron response in a postoperative nausea and vomiting setting, which was supported by a meta-analysis. We also highlight the heterogeneity and limitations of included studies as well as provide future directions for pharmacogenomics research.

PMID:39899439 | DOI:10.1111/cts.70108

ACS Synth Biol. 2025 Feb 3. doi: 10.1021/acssynbio.4c00853. Online ahead of print.

ABSTRACT

Yarrowia lipolytica has been widely engineered as a eukaryotic cell factory to produce various important compounds. However, the difficulty of gene editing and the lack of efficient neutral sites make rewiring of Y. lipolytica metabolism challenging. Herein, a Cas9 system was established to redesign the Y. lipolytica homologous recombination system, which caused a more than 56-fold increase in the HR efficiency. The fusion expression of the hBrex27 sequence in the C-terminus of Cas9 recruited more Rad51 protein, and the engineered Cas9 decreased NHEJ, achieving 85% single-gene positive efficiency and 25% multigene editing efficiency. With this system, neutral sites on different chromosomes were characterized, and a deep learning model was developed for gRNA activity prediction, thus providing the corresponding integration efficiency and expression intensity. Subsequently, the tool and platform strains were validated by applying them for the de novo synthesis of (S)-reticuline and (2S)-taxifolin. The developed platform strains and tools helped transform Y. lipolytica into an easy-to-operate model cell factory, similar to Saccharomyces cerevisiae.

PMID:39899813 | DOI:10.1021/acssynbio.4c00853

Adv Sci (Weinh). 2025 Feb 3:e2413405. doi: 10.1002/advs.202413405. Online ahead of print.

ABSTRACT

Unexpected toxicity has become a significant obstacle to drug candidate development, accounting for 30% of drug discovery failures. Traditional toxicity assessment through animal testing is costly and time-consuming. Big data and artificial intelligence (AI), especially machine learning (ML), are robustly contributing to innovation and progress in toxicology research. However, the optimal AI model for different types of toxicity usually varies, making it essential to conduct comparative analyses of AI methods across toxicity domains. The diverse data sources also pose challenges for researchers focusing on specific toxicity studies. In this review, 10 categories of drug-induced toxicity is examined, summarizing the characteristics and applicable ML models, including both predictive and interpretable algorithms, striking a balance between breadth and depth. Key databases and tools used in toxicity prediction are also highlighted, including toxicology, chemical, multi-omics, and benchmark databases, organized by their focus and function to clarify their roles in drug-induced toxicity prediction. Finally, strategies to turn challenges into opportunities are analyzed and discussed. This review may provide researchers with a valuable reference for understanding and utilizing the available resources to bridge prediction and mechanistic insights, and further advance the application of ML in drugs-induced toxicity prediction.

PMID:39899688 | DOI:10.1002/advs.202413405