The identification of cell types is a basic step of pipelines for Single-Cell RNA sequencing (scRNA-seq) data analysis. However, unsupervised clustering of cells from scRNA-seq data has multiple challenges: high dimensionality, sparseness of the expression matrix, and technical noise that generates false zero entries. In this study, we introduce new algorithms for clustering scRNA-seq data. The first algorithm builds a k-MST graph from distances obtained directly from the input data without dimensionality reduction. The computation follows an iterative procedure of k steps, calculating the edges of minimum spanning trees over different subgraphs obtained by removing edges selected in previous iterations. The Louvain algorithm is executed on the k-MST graph for cell clustering. We also explored an alternative based on neural networks, using an autoencoder to learn the parameters of a Gaussian mixture model. Benchmark experiments show that the algorithms have competitive accuracy, compared to previous solutions. Sequencing depth, number of cells and tissue types have important effects on the performance of the algorithms. Further experiments with scRNA-data taken from a patient with refractory epilepsy show that the autoencoder model achieved the best accuracy for this dataset, and the k-MST was competitive among graph-based approaches.

Abstract Motivation In prokaryotic genomes, methylation is an important epigenetic modification that regulates the uptake of foreign DNA; it can also contribute to replication or virulence. We present MPore, a novel method for the database-driven detection of active methyltransferases and their associated target site recognition motifs from Nanopore R10 sequencing data of prokaryotic isolates. In contrast to existing methods, which typically start with the de novo identification of differentially methylated sequence motifs, MPore starts by identifying potential methyltransferase genes by homology search against REBASE; activity is then assessed through a regularized logistic regression model of observed genome-wide methylation patterns, integrating motif and genomic sequence context information. Results On two benchmarking datasets 10 bacterial monocultures and two H. pylori genomes with complex methylation patterns, MPore achieved a combined recall of 93% and a combined PPV of 96%, outperforming Nanomotif (81%/91%), Modkit (66%/4%), and Snappy (89%/50%). Further validation on a well-characterized dataset of Mycoplasma hominis isolates showed perfect agreement with wet-lab-based validation results and demonstrated that MPore could complement REBASE information by disambiguating the specific methylated base in a motif with multiple potential methylation sites. MPore automatically produces integrated visualizations of the identified methyltransferases and observed methylation patterns; the tool is implemented as a user-friendly Snakemake pipeline. Availability MPore is freely available under the MIT license at https://github.com/DiltheyLab/MPore.

The identification of robust transcriptomic biomarkers remains a key challenge in oncology. To tackle this problem, hybrid ensemble feature selection (HEFS) methods have been developed to improve the stability of gene signatures by combining multiple algorithms and data perturbations. However, their results are often difficult to explore, interpret and reuse. To bridge this gap, we developed THe Biom (TCGA HEFS Biomarkers), an interactive application for visualization and comparative analysis of gene signatures across tumor stages and cancer types. The platform enables users to examine cancer-specific biomarkers, track changes across disease progression, and highlight shared features among signatures. THe Biom was built using previous HEFS analyses of six TCGA cancers across stages I to IV, and additional signatures can be added by users.Availability and implementation: THe Biom is freely available online at https://thebiom.compbio.ulaval.ca/, or for local use along with source code and datasets at https://github.com/MilanPicard/the_biom.

MotivationIn the global effort to characterize biodiversity, short species-specific genomic sequences known as DNA barcodes enable fine-grained comparisons among organisms within the same kingdom of life. Although machine learning algorithms specifically designed for the analysis of DNA barcodes are becoming more popular, most existing methodologies rely on generic supervised training algorithms.ResultsWe introduce BarcodeBERT, a family of models tailored to biodiversity analysis and trained exclusively on data from a reference library of 1.5 M invertebrate DNA barcodes. We evaluate BarcodeBERT on taxonomic identification tasks against a spectrum of machine learning approaches, including supervised training of classical neural architectures and fine-tuning of general DNA foundation models. Our self-supervised pretraining strategies on domain-specific data outperform fine-tuned foundation models, especially in identification tasks involving lower taxa such as genera and species. Compared with BLAST, a widely used sequence-search tool, BarcodeBERT achieves comparable species-level classification accuracy while being 55× faster. Our analysis of masking and tokenization strategies also provides practical guidance for building customized DNA language models, emphasizing the importance of aligning model training strategies with dataset characteristics and domain knowledge.Availability and implementationThe code repository is available at https://github.com/bioscan-ml/BarcodeBERT.

Network theory has proven invaluable in unraveling complex protein interactions. Previous studies have employed statistical methods rooted in network theory, including the Gaussian graphical model, to infer networks among proteins, identifying hub proteins based on key structural properties of networks such as degree centrality. However, there has been limited research examining a prognostic role of hub proteins on outcomes, while adjusting for clinical covariates in the context of high-dimensional data. To address this gap, we propose a network-guided penalized regression method. First, we construct a network using the Gaussian graphical model to identify hub proteins. Next, we preserve these identified hub proteins along with clinically relevant factors, while applying adaptive Lasso to non-hub proteins for variable selection. Our network-guided estimators are shown to have variable selection consistency and asymptotic normality. Simulation results suggest that our method produces better results compared to existing methods and demonstrates promise for advancing biomarker identification in proteomics research. Lastly, we apply our method to the Clinical Proteomic Tumor Analysis Consortium (CPTAC) data and identified hub proteins that may serve as prognostic biomarkers for various diseases, including rare genetic disorders and immune checkpoint for cancer immunotherapy. R package is freely available on CRAN repository (https://CRAN.R-project.org/package=NetGreg) and published under General Public License version 3.

Minimal residual disease (MRD) assessment has become a powerful tool in modern cancer management, offering early, actionable insights across diverse malignancies and often anticipating clinical events traditionally used to gauge therapeutic response. Beyond quantifying residual disease burden, MRD monitoring enables earlier intervention, nuanced treatment decisions, and improved outcome prediction before conventional endpoints appear. MRD negativity is now recognized as a potential surrogate endpoint, informing regulatory frameworks and accelerating drug development, as reflected by the FDA's approval of MRD as an accelerated endpoint in Multiple Myeloma trials. In particular, longitudinal MRD assessments, which track dynamic changes over time rather than single time points, have shown enhanced prognostic value, and more trials are now collecting such data. Building on this progress, we present MRDviz, an integrated R Shiny platform for exploratory analysis and simulation of longitudinal MRD data alongside survival outcomes. MRDviz offers interactive visualizations of MRD trajectories, real-time data quality checks, and identification of patient subgroups with distinct outcomes. Through coordinated views, customizable filtering, and integrated survival analyses, MRDviz empowers researchers to refine clinical hypotheses, design robust studies, and enhance the interpretation of MRD data, ultimately supporting better clinical decision-making and patient care. MRDviz is freely available at https://github.com/abbvie-external/MRDviz.

Single-cell RNA sequencing has substantially advanced our understanding of gene expression dynamics and cellular heterogeneity. In recent years, deep learning (DL) has emerged as a promising approach to infer genetic regulation. However, these methods still face challenges in representing complex regulatory mechanisms. Thus, it remains imperative to develop new algorithms to enhance both effectiveness and reliability. We propose DeepCE, a DL framework for correlation-enhanced gene regulatory network (GRN) inference. DeepCE strengthens the extraction of dynamic regulation by integrating bidirectional gated recurrent units with convolutional neural networks (CNNs). Specifically, bidirectional gated recurrent units captures dynamic temporal dependencies, while CNNs focuses on local spatial patterns within single-cell data, enabling the model to uncover complex gene-gene interactions and generate high-quality GRNs. This framework improves the accuracy and robustness of GRN inference by smoothing noisy gene expression data, extracting time-lagged regulatory signals, and filtering out spurious correlations. Experiments conducted on mouse and human datasets demonstrate the strong performance of DeepCE. Performance evaluations show that DeepCE outperforms existing methods, achieving the highest AUROC and AUPR scores. Codes for DeepCE are free available in the GitHub https://github.com/sxiaodai/DeepCE.

MotivationThe advent of ribosome profiling (an adaptation of RNA sequencing) to determine the translatome, has led to a huge improvement in our understanding of what parts of the transcriptome are translated. Many alternative open reading frames (ORFs) are now regularly being detected such as out-of-frame, overlapping, upstream or downstream reading frames, and alternative reading frames using non-canonical start codons. Various tools have been developed for the detection of such novel ORFs, but they lack the capacity to visually inspect reads—an important aspect of validation and prediction of translation.ResultsThe integrated and visualisation of ribosome profiling and RNA sequencing reads enables discrimination between transcriptional and translational signals, facilitating validation of predicted novel open reading frames. Furthermore, the inclusion of complementary evidence such as proteomic and long-read sequencing enables further validation of predicted novel open reading frames.Availability and implementationHere, we present, InspectorORF (https://www.github.com/aylz83/inspectorORF), an R package that readily plots ribosome profiling reads, alongside RNA sequencing reads across transcripts and/or ORFs. Additionally, custom information can be plotted including data from additional conditions and samples, proteomic analyses and reads from long-read sequencing.

Biological knowledgebases facilitate discovery across the life sciences by structuring experimental findings into human-readable and computable formats. These essential resources are maintained by a small number of professional biocurators worldwide and face combined chronic underfunding and the exponential growth of the literature. In this perspective, we review how artificial intelligence, particularly large language models and agentic systems, can augment literature-curation workflows. Applications include literature recommendation, entity recognition, data extraction, summarization, ontology development, and quality control with emphasis on published use cases at Global Core BioData Resources and ELIXIR Core Data Resources. We identify key challenges, including the scarcity of training data, difficulty in extracting complex relationships, and concerns about error propagation. To address these challenges, we propose a human-in-the-loop framework where generative artificial intelligence approaches accelerate routine tasks while curators provide critical evaluation and domain expertise. We also propose practical recommendations for the community, including the creation of shared benchmark datasets, harmonized evaluation frameworks, and best-practice guidelines for transparent human-in-the-loop AI deployment in biocuration. These synergistic partnerships will be critical to ensure biological rigour, accelerating knowledge integration while maintaining the quality essential for trusted biological resources.

Quality assessment and assembly comparison are essential steps while assembling new genomes. Many tools for evaluating assemblies typically provide synthetic parameters representing assembly quality or overall features, while others provide long detailed files where it is not always easy to identify and visualize the regions of correspondence and difference among different chromosome assemblies. Here we present ChromoMapper, a new tool which scans the output from QUAST, as well as other similar alignment description files, to quickly identify and display similarities and differences between the compared assemblies. It uses the information provided about aligned blocks, combined with additional annotations, to represent the main alignment regions at chromosomal or sub-chromosomal scale, highlighting similarities and collinearity between compared sequences, points of inconsistency, discontinuities, repeated regions and interruptions in the assembled sequences. ChromoMapper is available at https://chromomapper.ceinge.unina.it/ and via Zenodo (https://doi.org/10.5281/zenodo.16778863).