Bioinformatics (Oxford, England)[JOURNAL]
Sun
200 papers
EAGP: an efficient generative augmentation framework for phage protein classification under severe class imbalance.
Li J, Li H, Wang Y … +2 more , Zou Q, Zhu H
Bioinformatics · 2026 Jun · PMID 42296384 · Full text
MOTIVATION: The accurate classification of phage proteins is critical for advancing bacteriophage research. Despite the proliferation of machine learning approaches in this domain, the persistent issue of data imbalance...
MOTIVATION: The accurate classification of phage proteins is critical for advancing bacteriophage research. Despite the proliferation of machine learning approaches in this domain, the persistent issue of data imbalance continues to hinder performance, particularly for rare protein sequences. Previous attempts to address this by re-weighting minority classes have faced limitations due to insufficient feature extraction capabilities. RESULTS: In this paper, we introduce EAGP, a novel approach that integrates a generative model-functionally equivalent to a WGAN yet tailored for one-dimensional data-with the Evolutionary Scale Modeling (ESM) protein large language model for robust feature extraction. EAGP exhibits exceptional performance in binary classification and protein function annotation tasks. Crucially, our method not only improves overall classification efficacy but also significantly alleviates the performance degradation typically observed in minority classes. AVAILABILITY AND IMPLEMENTATION: The data and code underlying this article are available in GitHub at https://github.com/Innerly/EAGP and have been archived on Zenodo at https://doi.org/10.5281/zenodo.19928069.
Causal circuit tracing reveals distinct computational architectures in single-cell foundation models: inhibitory dominance, biological coherence, and cross-model convergence.
Kendiukhov I
Bioinformatics · 2026 Jun · PMID 42296381 · Publisher ↗
MOTIVATION: Sparse autoencoders (SAEs) decompose foundation-model activations into interpretable features, but the model-internal causal interactions between those features (i.e., what ablating one feature does to the ot...
MOTIVATION: Sparse autoencoders (SAEs) decompose foundation-model activations into interpretable features, but the model-internal causal interactions between those features (i.e., what ablating one feature does to the others, as distinct from the biological causal structure of the underlying cells)-and how those model-internal relationships relate to biological structure-are uncharacterised in single-cell foundation models (scFMs). RESULTS: We introduce model-internal causal circuit tracing-zeroing one SAE feature at a source layer and measuring the resulting change in all downstream SAE features, for each of 120 source features-and apply it to Geneformer V2-316M and scGPT whole-human across four conditions (96,892 ablation-derived edges, 80,191 forward passes). On annotation-selected source features, edges share GO/KEGG/Reactome/STRING/TRRUST ontology terms at 50.9-68.5%, a 2.9-6.2× enrichment over a configuration-preserving permutation null (p<0.002); on 20 randomly sampled source features this attenuates to 21.5-26.3%-still 2.5-3.1× above null-quantifying the annotation-selection contribution. Inhibitory dominance (fraction of ablation edges with d<0, i.e. source activation supports downstream target) is 65.5-89.4%. scGPT produces larger raw per-edge effects (mean |d|=1.40 vs. 1.05); after feature-share normalisation, Geneformer is stronger (paired gene-pair ratio 0.64 on 33,301 shared pairs). Cross-model consensus yields 1,142 architecture-invariant domain pairs (ordered pairs of GO biological-process categories "A→B" each connected by at least one ablation edge in both models; 10.6× enrichment over permutation null; p<0.001). Circuit edge magnitude explains <1% of the variance in marginal driver-gene co-expression on the same cells (R2=0.010, n=31,176): the graph encodes structure beyond bivariate correlation. Against a matched-cell-type ENCODE ChIP-seq prior, circuit-predicted TF→target pairs are enriched 2.06× (Fisher OR 5.84), markedly higher than 1.12× against TRRUST; direct ChIP-seq-supported target pairs show 10-30× larger CRISPRi sign-bias-corrected excess than indirect pairs. Gene-level CRISPRi validation on Replogle K562 and the non-cancer RPE1 arm (and a true primary-T-cell control from Shifrut 2018) after sign-bias correction shows excess over baseline of +0.03 and +0.35 percentage points on K562 and RPE1 respectively (baseline already 52-56% from sign marginals); effect-magnitude Spearman correlations ρ≈0. Bootstrap and per-cell-type stability (N∈{50,100,200{; B cell, CD4+ T, macrophage) give Pearson r≥0.97 on shared edges with 100% sign agreement; edge Jaccard grows monotonically with sample size. The circuit graph is therefore highly reproducible as an effect-size map, cell-type-specific in edge identity, consistent with co-expression encoding, and weakly-but-detectably enriched for ChIP-seq-supported direct regulatory edges.
HyperSeg-DG: multi-scale hyper feature context for domain-generalized medical image segmentation.
Islam MA, Rakib MYK, Huang Z … +2 more , XingFu W, Du W
Bioinformatics · 2026 Jun · PMID 42296380 · Full text
MOTIVATION: Developing segmentation models that remain reliable across diverse medical imaging domains and accurately delineate complex anatomical boundaries remains a persistent challenge for clinical deployment. Variat...
MOTIVATION: Developing segmentation models that remain reliable across diverse medical imaging domains and accurately delineate complex anatomical boundaries remains a persistent challenge for clinical deployment. Variations in imaging modalities, scanners, and acquisition settings introduce significant domain shifts, while fuzzy or overlapping tissue boundaries further complicate precise segmentation. Despite extensive research, most approaches address these challenges separately, leading to limited generalization and reduced robustness in real-world clinical scenarios. RESULTS: To overcome these limitations, we propose HyperSeg-DG, a novel medical image segmentation approach that integrates the WMamba backbone with the Multi-Scale Hyper Feature Context Block (HFCB). The HFCB addresses foreground-background uncertainty and boundary ambiguities by capturing multi-scale feature relations and long-range contextual dependencies. This enables the model to focus on relevant pathological features while helping reduce the influence of irrelevant co-occurring ones, such as similarly sized polyps, especially in low-contrast or poorly lit environments. WMamba further improves domain generalization by processing images in localized windows and using its selective 2D scanning mechanism to learn robust, transferable features that reduce feature misalignment under domain shift. Extensive experiments across multiple medical segmentation benchmarks demonstrate that HyperSeg-DG achieves consistent 2%-3% improvements over strong baselines, confirming its effectiveness in enhancing segmentation performance and generalization across diverse, unseen domains. AVAILABILITY AND IMPLEMENTATION: The code and datasets of HyperSeg-DG are available at https://github.com/Pollob001/HyperSeg-DG.
SECTOR: structural entropy-based learning of spatiotemporal organisation in spatial transcriptomics.
Huang L, Zhang J, Gong W … +3 more , Zeng G, Peng H, Chen D
Bioinformatics · 2026 Jun · PMID 42296338 · Full text
MOTIVATION: Spatial transcriptomics (ST) profiles gene expression in tissue context, enabling spatial domain detection. However, relatively few methods jointly recover discrete spatial domains and continuous within-secti...
MOTIVATION: Spatial transcriptomics (ST) profiles gene expression in tissue context, enabling spatial domain detection. However, relatively few methods jointly recover discrete spatial domains and continuous within-section pseudotemporal trends in a single framework. Current spatiotemporal approaches often emphasise trajectory continuity to recover smooth progression-associated gradients, but this may blur neighbouring domain boundaries and reduce clustering accuracy. Conversely, specialised spatial clustering algorithms typically rely on external single-cell trajectory tools rather than providing an integrated, spatially aware pseudotime model. RESULTS: We introduce SECTOR (Structural Entropy-based Clustering and pseudoTime ORdering), a lightweight deep graph learning framework that unifies spatial domain detection and pseudotime inference. SECTOR optimises a differentiable structural entropy (SE) objective on a fused spatial-expression graph, with spatial total variation regularisation to promote tissue continuity. Across seven benchmark datasets spanning standard and modern high-resolution ST platforms, SECTOR consistently outperformed existing spatiotemporal methods in clustering accuracy and matched or exceeded leading spatial clustering algorithms, while maintaining modest computational demands. In human breast cancer and mouse olfactory bulb case studies, SECTOR recovered spatially organised pseudotime patterns supported by semivariance, transition-gene, enrichment and marker-gene analyses. Together, these results show that SE-based learning provides an effective and scalable strategy for modelling within-section spatiotemporal organisation in ST. AVAILABILITY: SECTOR is available on GitHub at https://github.com/lhbcb/SECTOR and archived on Figshare at https://doi.org/10.6084/m9.figshare.32029830.
A survey of models composed of graph neural networks and large language models for molecular science.
Segura-Alabart N, Serratosa F
Bioinformatics · 2026 Jul · PMID 42296336 · Full text
MOTIVATION: Graphs have been demonstrated to have an impressive ability to keep the structural and semantic properties of chemical compounds and are therefore widely used in molecular modeling. Moreover, Graph Neural Net...
MOTIVATION: Graphs have been demonstrated to have an impressive ability to keep the structural and semantic properties of chemical compounds and are therefore widely used in molecular modeling. Moreover, Graph Neural Networks (GNNs) and Large Language Models (LLMs) have achieved outstanding results in predicting molecular properties and generating textual descriptions given graphs or texts, respectively. Recently, some mathematical models have been presented for molecular science applications, which combine the GNNs ability to capture the structural and semantic information and the generative ability of LLMs. However, these models are dispersed across the literature and vary in architecture, input-output design, and application scope, making it difficult to systematically compare them or to identify suitable approaches for specific research objectives. RESULTS: This paper classifies recent GNN-LLM models and summarizes them with the aim of providing guidance for research involving their use or the development of new models. We present tables that depict the properties and parameters of the models, as well as their associated chemical computational applications. In addition, a new model classification is presented to help researchers define the models they use or future ones. Finally, we quantitatively compare these models based on their reported experimental results.
DrugDL: dual-modal deep learning framework for multi-property drug prediction and targeted therapy discovery.
Zhang Q, Yu X, Wei Y … +5 more , Xia Y, Shen LC, Wang ZH, Shen HB, Yu DJ
Bioinformatics · 2026 Jul · PMID 42296331 · Full text
MOTIVATION: The accurate and robust representation of drug molecule features, the prediction of drug-target biomacromolecule interactions, and the determination of physicochemical properties are crucial in drug developme...
MOTIVATION: The accurate and robust representation of drug molecule features, the prediction of drug-target biomacromolecule interactions, and the determination of physicochemical properties are crucial in drug development. However, these tasks remain challenging due to issues such as the limited generalizability of single-modal representations, the absence of multitask prediction frameworks, and weak adaptability in cold-start scenarios. RESULTS: In this study, we present DrugDL, a framework for comprehensive drug molecule representation and the prediction of multiple downstream tasks, including drug-target interactions, binding affinities, binding sites, physicochemical properties, toxicity, and drug-drug interactions. DrugDL jointly learns representations of the drug chemical space and the target protein biological space, while capturing multiscale interaction mechanisms between drug molecules and target proteins through the integration of cross-modal contrastive learning and single-modal feature enhancement algorithms. Specifically, DrugDL employs a multitask prediction framework to predict multiple properties of drug molecules. In practical applications, it consistently outperforms state-of-the-art methods, particularly in cold-start tasks. The framework has been successfully applied to high-throughput screening, the identification of inhibitors of SARS-CoV-2 and metabolic enzymes, and the prediction of cancer-targeted drugs. Experimental validations on EGFR and ALK targets further demonstrate its effectiveness as a precise drug discovery tool. By enabling accurate molecular representation and multi-property prediction, DrugDL provides end-to-end technical support for drug development, thereby significantly accelerating the drug discovery process. AVAILABILITY AND IMPLEMENTATION: The datasets and code are available at https://github.com/ZhangQi9910/DrugDL. The version of record is archived in Zenodo with the DOI: 10.5281/zenodo.20579718.
LinearCapR: linear-time computation of per-nucleotide structural-context probabilities of RNA without base-pair span limits.
Otagaki T, Hosokawa H, Fukunaga T … +3 more , Iwakiri J, Terai G, Asai K
Bioinformatics · 2026 Jun · PMID 42295831 · Full text
MOTIVATION: RNA molecules adopt dynamic ensembles of secondary structures, where the local structural context of each nucleotide-such as whether it resides in a stem or a specific type of loop-strongly shapes molecular i...
MOTIVATION: RNA molecules adopt dynamic ensembles of secondary structures, where the local structural context of each nucleotide-such as whether it resides in a stem or a specific type of loop-strongly shapes molecular interactions and regulatory function. Structural-context probabilities therefore provide a more functionally informative view of RNA folding than the minimum free energy structures or base-pairing probabilities. However, existing tools either require O(N3) time or employ span-restricted approximations that omit long-range base-pairs, limiting their applicability to large and biologically important RNAs. RESULTS: We introduce LinearCapR, enabling linear-time, span-unrestricted computation of structural-context marginalized probabilities, using beam-pruned Stochastic Context Free Grammar-based computation. LinearCapR retains global ensemble features lost by span-limited methods and yields superior predictive power on bpRNA-1m(90) dataset, especially for multiloops and exterior regions, as well as long-distance stems. LinearCapR supports analysis of long RNAs, demonstrated on the full genome of SARS-CoV-2. LinearCapR provides the first base-pair-span-unrestricted, linear-time framework for RNA structural-context analysis, retaining key thermodynamic ensemble features essential for functional interpretation. It enables large-scale studies of viral genomes, long non-coding RNAs, and downstream analyses such as RNA-binding protein site prediction. AVAILABILITY AND IMPLEMENTATION: The source code of LinearCapR is available at https://github.com/hoget157/LinearCapR. The archived software release used in this work is available at Zenodo: https://doi.org/10.5281/zenodo.19450645.
spAttClu: a spatial domain clustering model leveraging spatially weighted graph attention and contrastive learning.
Zhang T, Zhang R, Zhang H … +6 more , Zhao Z, Wang R, Li S, Jiang Y, Wei B, Wang G
Bioinformatics · 2026 Jun · PMID 42294550 · Full text
MOTIVATION: The rapid growth of spatial transcriptomics data holds potential for deep understanding of spatial specificity and tissue heterogeneity. Recognizing spatial domains is a fundamental step for deciphering tissu...
MOTIVATION: The rapid growth of spatial transcriptomics data holds potential for deep understanding of spatial specificity and tissue heterogeneity. Recognizing spatial domains is a fundamental step for deciphering tissue functional architecture and dissecting tissue heterogeneity. However, existing models typically define adjacency relations using static weights, which cannot dynamically adjust neighbor importance based on expression context, thereby limiting the accuracy and robustness of spatial domain recognition. RESULTS: We propose spAttClu, a clustering model integrating spatially weighted graph attention with contrastive learning. It adaptively learns neighbor contributions in varying contexts through a distance-weighted graph attention mechanism and enhances embedding discriminability via multi-level contrastive learning. spAttClu demonstrates superior clustering performance on the DLPFC dataset. Moreover, it shows cross-platform adaptability and enables vertical/horizontal inte-gration of multiple tissue slices.
PMGen: from peptide-MHC structure prediction to peptide generation.
Asgary AH, Aleyasin A, Mehl JA … +6 more , Fallah SS, Aintablian H, Ludewig B, Mishto M, Liepe J, Söding J
Bioinformatics · 2026 Jun · PMID 42289978 · Full text
MOTIVATION: Accurate structural modeling of peptide-major histocompatibility complex (pMHC) complexes is essential for structure-driven immunotherapy design, yet current prediction tools suffer from narrow class coverage...
MOTIVATION: Accurate structural modeling of peptide-major histocompatibility complex (pMHC) complexes is essential for structure-driven immunotherapy design, yet current prediction tools suffer from narrow class coverage, restricted peptide lengths, insufficient accuracy, and a lack of built-in structure-aware peptide sampling. Consequently, most mimotope and altered peptide ligand designs rely solely on sequence substitution, leaving spatial and biophysical insights from pMHC structures largely unexploited. RESULTS: We introduce peptide-MHC generator (PMGen), an integrated framework for structure prediction and structure-guided design of variable-length peptides across MHC Class I and II. PMGen enforces anchor constraints within AlphaFold2 through two complementary strategies, initial guess and template engineering, achieving state-of-the-art structural fidelity without model fine-tuning. On a comprehensive benchmark, PMGen outperforms all existing methods, yielding median peptide-core Cα RMSDs of 0.62 Å for MHC-I and 0.33 Å for MHC-II. We show that PMGen can recover incorrectly predicted anchor positions and that AlphaFold pLDDT scores enable sequence-independent binding-core identification. Applied to a published neoantigen/wild-type pair, PMGen accurately captures mutation-induced conformational changes. Beyond structure prediction, we show that ProteinMPNN sampling on PMGen-predicted backbones yields higher affinity peptides while preserving the parental 3D conformation. Using PMGen to generate 63 817 high-confidence pMHC structures as training data, we further improve ProteinMPNN's peptide sequence recovery from 0.14 to 0.64 on a test set of 85 unseen MHC-I alleles, highlighting the value of accurate predicted structures for downstream machine learning tasks. AVAILABILITY AND IMPLEMENTATION: PMGen is freely available at https://github.com/soedinglab/PMGen, with an interactive Colab notebook at https://colab.research.google.com/github/soedinglab/PMGen/blob/master/colab.ipynb.
GBSC: Graph-Based Sequence Clustering method for similar short tandem repeats in protein sequences.
Jarnot P, Ziemska-Legiecka J, Grynberg M … +2 more , Promponas VJ, Gruca A
Bioinformatics · 2026 Jun · PMID 42289972 · Publisher ↗
MOTIVATION: Short tandem repeats (STRs) are abundant in protein sequences and play an important role in determining their structures and functions. Strikingly, the unusual compositional characteristics of tandem repeats...
MOTIVATION: Short tandem repeats (STRs) are abundant in protein sequences and play an important role in determining their structures and functions. Strikingly, the unusual compositional characteristics of tandem repeats break classical sequence analysis tools. RESULTS: We here establish the first algorithm to effectively identify and cluster STRs: Graph-Based Sequence Clustering (GBSC) features linear time complexity, and clusters protein sequence fragments based on their STRs, while allowing for insertions and mutations, supporting an analysis of imperfect or cryptic repeats. Due to its computational efficacy, our algorithm can be used to systematically scan for patterns in large datasets. We compare our method both with state-of-the-art methods for identifying STRs in proteins and alternative clustering approaches. Unlike existing STR analysis methods, GBSC clusters repeat patterns rather than raw sequences, operating at the level of structural repeat identity, while tolerating biological variations and preventing erroneous merging of structurally and functionally distinct motifs. Whereas functional annotation is typically only available at the protein level, the functions of individual STRs and sequences of adjacent STRs remains largely unknown. On a challenging use-case we here demonstrate and discuss how our method can be used to associate previously unannotated repetitive protein fragments with similar ones, allowing a transfer of annotation by similarity. For the first time, GBSC offers a tool that systematically extends this fundamental bioinformatics principle to low-complexity regions across large datasets. AVAILABILITY AND IMPLEMENTATION: GBSC is available at GitHub https://github.com/patryk-jarnot/GBSC and https://doi.org/10.5281/zenodo.18965247. The data and scripts to reproduce the analysis are available at https://doi.org/10.5281/zenodo.16906653.
SPPIDER-seq: sequence-based partner-aware predictor of protein-protein interaction sites.
Porollo A, Jadhav O, Alvarez A … +1 more , Chen J
Bioinformatics · 2026 Jul · PMID 42289971 · Full text
MOTIVATION: Sequence-based protein-protein interaction (PPI) site predictors typically analyze proteins in isolation, neglecting partner-specific context that is critical for interface specificity, particularly in transi...
MOTIVATION: Sequence-based protein-protein interaction (PPI) site predictors typically analyze proteins in isolation, neglecting partner-specific context that is critical for interface specificity, particularly in transient and disordered interactions. RESULTS: We introduce SPPIDER-seq, a partner-aware PPI site prediction framework that combines pretrained ESM-2 embeddings with a cross-attention architecture to enable residue-level conditioning on interacting partners. We curated non-redundant protein-peptide interaction datasets from BioLiP and used them to train and benchmark two complementary models: a receptor-centric model optimized for structured interfaces and a peptide-centric model tailored to disordered, motif-driven binding. On blind benchmarks, SPPIDER-seq achieved AUROC values up to 0.797 and MCC values up to 0.269, outperforming AlphaFold3 on peptide-mediated and disordered interfaces while remaining complementary on globular complexes. Application to 341 TP53 interaction partners revealed coherent, partner-specific interface patterns across both structured and intrinsically disordered regions. AVAILABILITY: SPPIDER-seq models, datasets, and the Python code are freely available on the web at https://github.com/aporollo-lab/SPPIDER-seq and archived on Zenodo at DOI: 10.5281/zenodo.19835990, corresponding to GitHub release v2.0-manuscript.
GRNContext: An Interactive Web Platform for Contextualized Gene Regulatory Networks Visualization Across Human Cancers.
Pezoa-Soto I, Hernández-Galaz S, de Las Rivas J … +3 more , Lladser A, Varas-Godoy M, Martin AJM
Bioinformatics · 2026 Jun · PMID 42289970 · Publisher ↗
SUMMARY: While current Gene Regulatory Network (GRN) databases provide comprehensive reference maps of potential interactions between transcription factors and target genes, they do not specify which regulatory interacti...
SUMMARY: While current Gene Regulatory Network (GRN) databases provide comprehensive reference maps of potential interactions between transcription factors and target genes, they do not specify which regulatory interactions are active within specific biological contexts. This limitation is particularly critical in cancer, where transcriptional programs are inherently tissue-specific. To address this gap, we developed GRNContext, an interactive web platform designed for the visualization, exploration, and comparative analysis of gene regulatory networks contextualized across 33 cancer types from The Cancer Genome Atlas (TCGA). Our approach uses the TFLink human reference GRN as a starting point and integrates TCGA transcriptomic profiles to infer cancer-specific regulatory activity. Regulatory relevance was assessed using complementary machine learning and statistical methods, which were unified into a consensus score to prioritize and filter the most relevant candidate regulators for each target gene. By providing both curated context-specific GRNs and a user-friendly platform, GRNContext constitutes a comprehensive and accessible resource that supports mechanistic investigations, hypothesis generation, and translational research focused on transcriptional regulation in cancer. AVAILABILITY AND IMPLEMENTATION: GRNContext is supported by all major browsers and freely available on the web at https://apps.cienciavida.org/grncontext. It is implemented as a client-server web application featuring a FastAPI backend and a React frontend utilizing Cytoscape.js for interactive network visualization, all containerized via Docker for cross-platform compatibility. SUPPLEMENTARY INFORMATION: The source code, supplementary materials, and resources for the GRNContext web platform are accessible at https://doi.org/10.5281/zenodo.18355998..
CistromeMeta: a large language model powered tool for automated ChIP-seq metadata extraction.
Piccaro N, Brown M, Meyer C
Bioinformatics · 2026 Jun · PMID 42287723 · Full text
SUMMARY: Public repositories such as NCBI's Gene Expression Omnibus (GEO) contain large numbers of ChIP-seq experiments, but their reuse is limited by heterogeneous free-text metadata describing target proteins, histone...
SUMMARY: Public repositories such as NCBI's Gene Expression Omnibus (GEO) contain large numbers of ChIP-seq experiments, but their reuse is limited by heterogeneous free-text metadata describing target proteins, histone marks, cell lines, tissues, and disease states. We introduce CistromeMeta, a Python-based command-line tool that leverages large language models (LLMs) in a few-shot setting to automatically extract and standardize ChIP-seq metadata from GEO XML records without custom model training. The tool validates extracted terms against authoritative biological databases, including NCBI Gene, Harmonizome 3.0, AnimalTFDB 4.0, Cellosaurus, Experimental Factor Ontology, and Uberon, producing standardized outputs with official gene symbols and ontology identifiers for scalable metadata curation. AVAILABILITY AND IMPLEMENTATION: The Python source code is freely available at https://github.com/nickpiccaro/CistromeMetaX. An archived version of the software is available through Zenodo at DOI: 10.5281/zenodo.20244834. The tool requires Python 3.6+ and an OpenAI API key.
PEPE: scalable extraction of multi-modal protein language model representations.
Zhong J, Cardente N, Sandve GK … +3 more , Bashour H, Abbate MF, Greiff V
Bioinformatics · 2026 Jul · PMID 42286792 · Full text
SUMMARY: Protein language models (PLMs) capture intricate amino-acid dependencies, producing embeddings that encode rich structural, functional, and evolutionary information. Despite their potential, current extraction w...
SUMMARY: Protein language models (PLMs) capture intricate amino-acid dependencies, producing embeddings that encode rich structural, functional, and evolutionary information. Despite their potential, current extraction workflows rely on arbitrary choices, with respect to embedding layer, pooling, and padding, that frequently yield suboptimal representations for feature extraction and downstream analyses. Large-scale embedding generation is further limited by inefficiencies in computation and memory: (i) accumulating all model outputs in memory before writing to disk causes severe bottlenecks, and (ii) repeatedly embedding identical sequences to extract different modes introduces redundant computation and drastically reduces throughput and scalability. We introduce PEPE (Parallel Extraction for Protein Embeddings), a command-line tool and Python library that enables efficient, high-throughput, and multimodal extraction from protein language models. PEPE's parallelized and streaming-based architecture achieves runtimes several orders of magnitude faster than sequential approaches. Unlike conventional methods-whose peak memory usage scales linearly with output size and fails when memory capacity is exceeded-PEPE maintains stable, low memory consumption, enabling multimodal embedding extraction even beyond available RAM. PEPE supports a wide range of state-of-the-art and custom PLMs through a simple, flexible interface. By combining scalability, robustness, and ease of use, PEPE allows researchers to generate massive, information-rich embedding datasets efficiently, and facilitate the discovery of optimal representations for structural, functional, and evolutionary downstream tasks. By streamlining the generation of diverse embedding configurations, PEPE provides researchers with the necessary data to identify high-performing latent states for specific biological contexts without requiring additional computational resources. AVAILABILITY AND IMPLEMENTATION: PEPE is a command-line tool written in Python and published under MIT license. The source code and documentation are available at https://github.com/csi-greifflab/pepe-cli. PEPE is also available for installation from PyPI under https://pypi.org/project/pepe-cli and deposited on Zenodo at https://zenodo.org/records/20268104.
Bayesian Hyperparameter Optimization Improves scGPT Fine-Tuning for Single-Cell Multi-Omics Integration.
Jun Tay DY, Khanh Le NQ, Heng Chua MC
Bioinformatics · 2026 Jun · PMID 42286785 · Publisher ↗
MOTIVATION: Foundation models such as scGPT have demonstrated strong potential for single-cell multi-omics integration; however, their downstream performance is highly sensitive to hyperparameter selection. Manual fine-t...
MOTIVATION: Foundation models such as scGPT have demonstrated strong potential for single-cell multi-omics integration; however, their downstream performance is highly sensitive to hyperparameter selection. Manual fine-tuning remains computationally expensive, dataset-dependent, and often irreproducible. Despite the increasing adoption of foundation models in single-cell analysis, systematic strategies for robust hyperparameter optimization remain underexplored. RESULTS: We developed a Bayesian optimization framework based on Tree-structured Parzen Estimators (TPE) for automated fine-tuning of scGPT and evaluated its performance on two benchmark bone marrow mononuclear cell (BMMC) multi-omics datasets, including CITE-seq and GSE194122 datasets. Across datasets, Bayesian optimization consistently improved biological conservation and batch integration metrics compared with default scGPT configurations. On the original BMMC benchmark, optimization improved AvgBIO from 0.59 to 0.67 and PCR from 0.33 to 0.52. On the GSE194122 dataset, the default configuration exhibited unstable convergence and weak biological preservation (AvgBIO = 0.19; ARI = 0.007), whereas Bayesian optimization substantially improved integration performance (AvgBIO = 0.60; ARI = 0.63) while reducing validation loss from 137 to 47.1. These findings demonstrate substantial dataset-specific sensitivity of scGPT fine-tuning and highlight the importance of automated optimization for stable deployment across heterogeneous multi-omics datasets. CONCLUSION: Our study demonstrates that Bayesian optimization provides an effective and reproducible strategy for stabilizing scGPT fine-tuning across diverse single-cell multi-omics datasets. Rather than introducing a new integration architecture, this work emphasizes the importance of systematic optimization for improving robustness and reproducibility of foundation-model applications in computational biology. AVAILABILITY: Our model and dataset are freely available at: https://github.com/daren642/scGPT_multiomic_tuning.
SEMPLR: an R package for transcription factor binding prediction.
Kenney GE, Sherpa RN, Burgess JD … +2 more , Boyle AP, Phanstiel DH
Bioinformatics · 2026 Jun · PMID 42286344 · Full text
SUMMARY: SEMPLR is an R package that predicts transcription factor binding and variant effects using SNP Effect Matrices (SEMs), providing efficient, genome-wide scoring, enrichment testing, and visualization tools for c...
SUMMARY: SEMPLR is an R package that predicts transcription factor binding and variant effects using SNP Effect Matrices (SEMs), providing efficient, genome-wide scoring, enrichment testing, and visualization tools for comprehensive analysis of regulatory sequences. AVAILABILITY: Available on GitHub at https://github.com/grkenney/SEMPLR and on Bioconductor at https://bioconductor.org/packages/release/bioc/html/SEMPLR.html.
Harmonization and integration of pharmacogenomics screens.
Chen AT, Kelly MR, Ideker T … +1 more , Mattson NM
Bioinformatics · 2026 Jul · PMID 42286248 · Full text
MOTIVATION: Large pharmacogenomics screens have generated a wealth of information cataloguing the responses of more than a thousand tumor cell-line models to FDA-approved and exploratory drugs. Although centralized repos...
MOTIVATION: Large pharmacogenomics screens have generated a wealth of information cataloguing the responses of more than a thousand tumor cell-line models to FDA-approved and exploratory drugs. Although centralized repositories have consolidated data access, the diversity of experimental platforms and response metrics used in these screens have made it challenging to integrate and compare their measured drug responses. Towards better pharmacogenomic data harmonization, we surveyed a range of data analysis protocols based on different curve-fitting functions (sigmoid, piecewise linear), different response metrics (IC50, EC50, integrated AUC), and different drug concentration windows (full range or truncated). RESULTS: We found that an AUC derived from a sigmoidal curve fitted to a truncated dose range yields the strongest agreement between screening platforms, significantly bettering other protocols surveyed. This harmonization procedure also best aligns drug responses across successive iterations of the same platform. These findings broadly inform efforts to integrate drug response data in large-scale analyses. AVAILABILITY: The source code to generate drug response profiles and correlations are available at https://github.com/digitaltumors/Pharmacogenomics_Screens_Harmonization.git.
BioNeuralNet: a graph neural network based Multi-Omics network data analysis tool.
Ramos V, Hussein S, Abdel-Hafiz M … +6 more , Sarkar A, Liu W, Kechris KJ, Bowler RP, Lange L, Banaei-Kashani F
Bioinformatics · 2026 Jun · PMID 42274222 · Full text
SUMMARY: Multi-omics data offer unprecedented insights into complex biological systems, yet their high dimensionality, sparsity, and intricate interactions pose significant analytical challenges. Network-based approaches...
SUMMARY: Multi-omics data offer unprecedented insights into complex biological systems, yet their high dimensionality, sparsity, and intricate interactions pose significant analytical challenges. Network-based approaches have advanced multi-omics research by effectively capturing biologically relevant relationships among molecular features (e.g., genes, proteins, metabolites). While these methods are powerful for representing molecular interactions, there remains a need for tools specifically designed to effectively utilize these network representations across diverse downstream analyses. To fulfill this need, we introduce BioNeuralNet, a flexible and modular Python framework tailored for end-to-end network-based multi-omics data analysis. BioNeuralNet leverages Graph Neural Networks (GNNs) to learn biologically meaningful low-dimensional representations from multi-omics networks, converting these complex molecular networks into versatile embeddings. BioNeuralNet supports all major stages of multi-omics network analysis, including several network construction techniques, generation of low-dimensional representations, and a broad range of downstream analytical tasks. Its extensive utilities, including diverse GNN architectures, and compatibility with established Python packages (e.g., scikit-learn, PyTorch, NetworkX), enhance usability and facilitate quick adoption. BioNeuralNet is an open-source, user-friendly, and extensively documented framework designed to support flexible and reproducible multi-omics network analysis in precision medicine. AVAILABILITY AND IMPLEMENTATION: The BioNeuralNet library is available via The Python Package Index (PyPI). Source code, documentation, tutorials, and workflows are hosted at https://bioneuralnet.readthedocs.io. Code archived at https://doi.org/10.5281/zenodo.17503083.
LCR-modules: a collection of workflows for cancer genome analysis.
Dreval K, Hilton LK, Grande BM … +15 more , Banco G, Coyle KM, Cruz M, Gillis S, Klossok L, Pararajalingam P, Rushton CK, Shaalan H, Thomas N, Winata H, Wong J, Yiu J, Steidl C, Scott DW, Morin RD
Bioinformatics · 2026 Jun · PMID 42271624 · Full text
MOTIVATION: The surge of genomic data from advanced sequencing technologies is outpacing current analytical pipelines. We introduce LCR-modules, an open-source suite of bioinformatics tools designed for flexible and auto...
MOTIVATION: The surge of genomic data from advanced sequencing technologies is outpacing current analytical pipelines. We introduce LCR-modules, an open-source suite of bioinformatics tools designed for flexible and automated cancer genome data analysis. LCR-modules enables reproducible analysis of diverse cancer genomics data at scale. The suite comprises 49 Snakemake-based workflows organized into three levels, facilitating tasks from low-level quality control to complex cohort-level analyses. LCR-modules supports various sequencing types and integrates pipelines such as mutation calling, expression quantification, and cohort-level aggregation, ensuring flexibility and reproducibility. LCR-modules represents a significant advancement in genomic data analysis, reducing barriers in reproducibility and scalability and has already been applied to a combination of exomes and genomes from over 10 800 samples. AVAILABILITY: No new data were generated in support of this research. The source code for the LCR-modules is openly available at https://github.com/LCR-BCCRC/lcr-modules.
AEGIS: an annotation extraction and genomic integration resource.
Navarro-Payá D, Santiago A, Velt A … +3 more , Moretto M, Rustenholz C, Matus JT
Bioinformatics · 2026 Jun · PMID 42262904 · Full text
MOTIVATION: Genome annotation files (GFF3/GTF) are the standard for storing genomic feature data, yet their flexibility often results in formatting inconsistencies that create bottlenecks for downstream bioinformatics an...
MOTIVATION: Genome annotation files (GFF3/GTF) are the standard for storing genomic feature data, yet their flexibility often results in formatting inconsistencies that create bottlenecks for downstream bioinformatics analyses. A robust, unified framework is required to parse, standardise, and validate these files to ensure interoperability and facilitate complex comparative genomic tasks. RESULTS: We present AEGIS (Annotation Extraction and Genomic Integration Suite), a comprehensive toolkit designed to parse, correct, and standardise genome annotations. Beyond quality control, AEGIS provides advanced modules for flexible feature extraction (e.g., coding sequences, promoters) and comparative genomic analysis. Uniquely, it integrates multiple lines of evidence, including sequence homology, synteny, and coordinate-based lift-overs, to assess gene model correspondence and infer orthology. We demonstrate the utility of AEGIS by quantifying complex structural changes between Arabidopsis annotation versions and identifying high-confidence orthologues across diverse plant genomes. AVAILABILITY: AEGIS is implemented in Python. Source code and documentation are freely available under the GPL-3 license at https://github.com/Tomsbiolab/aegis and as a Docker container at https://hub.docker.com/r/tomsbiolab/aegis. The package is also available on PyPI (pip install aegis-bio).
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe