Bioinformatics (Oxford, England)[JOURNAL]
Sun
200 papers
CodonMoE: DNA language models for codon-dependent mRNA prediction.
Du S, Liang L, Li J … +1 more , Kingsford C
Bioinformatics · 2026 Jun · PMID 42108565 · Full text
MOTIVATION: Genomic language models (gLMs) face a fundamental efficiency challenge: one must either maintain separate specialized models for each biological modality (DNA and RNA) or develop large multimodal architecture...
MOTIVATION: Genomic language models (gLMs) face a fundamental efficiency challenge: one must either maintain separate specialized models for each biological modality (DNA and RNA) or develop large multimodal architectures. Both approaches impose significant computational burdens-modality-specific models require redundant infrastructure despite inherent biological connections, while multi-modal architectures demand increased parameter counts and extensive cross-modality pretraining. RESULTS: To address this limitation, we introduce CodonMoE (Adaptive Mixture of Codon Reformative Experts), a lightweight adapter that transforms DNA language models into effective RNA analyzers without RNA-specific pretraining. Our theoretical analysis establishes CodonMoE as a universal approximator at the codon level, capable of mapping arbitrary functions from codon sequences to codon-dependent RNA properties given sufficient expert capacity. Across four RNA prediction tasks spanning stability, expression, and regulation, DNA models augmented with CodonMoE significantly outperform their unmodified counterparts, with the HyenaDNA+CodonMoE series achieving state-of-the-art results using 80% fewer parameters than specialized RNA models. By maintaining sub-quadratic complexity while achieving superior performance, our approach provides a principled path toward unifying genomic language modeling, leveraging more abundant DNA data and reducing computational overhead while preserving modality-specific performance advantages. AVAILABILITY AND IMPLEMENTATION: Source code for the method and to reproduce the results is available at https://github.com/Kingsford-Group/CodonMoE.
PEARL: integrative multi-omics classification and omics feature discovery via deep graph learning.
Zhao Q, Du J, Zhou M … +3 more , Wang XW, Sun Q, Chen C
Bioinformatics · 2026 Jun · PMID 42108553 · Full text
MOTIVATION: Integrating multi-omics data provides valuable insights into biological processes by capturing information across multiple molecular layers, enabling a comprehensive understanding of complex diseases and driv...
MOTIVATION: Integrating multi-omics data provides valuable insights into biological processes by capturing information across multiple molecular layers, enabling a comprehensive understanding of complex diseases and driving advancements in precision medicine. However, existing computational methods for multi-omics integration face significant challenges, such as low reliability and poor generalizability, due to the high dimensionality and low sample size nature of omics data. RESULTS: To address these challenges, we present PEARL (Pearson-Enhanced spectrAl gRaph convoLutional networks), a novel deep graph learning method for biomedical classification and functional important omics features identification. PEARL leverages a simple yet effective learning architecture to achieve superior and robust performance in high-dimensional, low-sample-size multi-omics settings. Our results demonstrate that PEARL significantly outperforms existing state-of-the-art methods on both synthetic and real biomedical datasets. Furthermore, applied to Alzheimer's disease (AD) brain multi-omics data, features prioritized by PEARL lead to functionally important genes that demonstrate significant enrichment in AD-related pathways. These findings highlight PEARL's practical utility in biomedical research and its potential to enhance biological interpretability in multi-omics studies. AVAILABILITY AND IMPLEMENTATION: The source code of our computational framework is available at https://github.com/zqq121017/PEARL.
anndataR improves interoperability between R and Python in single-cell transcriptomics.
Deconinck L, Zappia L, Cannoodt R … +8 more , Morgan M, Virshup I, Sang-Aram C, Bredikhin D, Schilder B, Seurinck R, Saeys Y, scverse core
Bioinformatics · 2026 Jun · PMID 42108549 · Full text
SUMMARY: Many single-cell transcriptomics datasets are stored in the HDF5-backed AnnData (H5AD) file format, as popularized by the Python scverse ecosystem. However, accessing these datasets from R, allowing users to tak...
SUMMARY: Many single-cell transcriptomics datasets are stored in the HDF5-backed AnnData (H5AD) file format, as popularized by the Python scverse ecosystem. However, accessing these datasets from R, allowing users to take advantage of the strengths of each language, can be difficult. anndataR facilitates this access by allowing users to natively read and write H5AD files in R, convert them to and from SingleCellExperiment or Seurat objects, or even work with the resulting R AnnData object directly. We perform rigorous testing to ensure compatibility between Python-written and R-written H5AD files, guaranteeing long-term interoperability between languages. AVAILABILITY AND IMPLEMENTATION: anndataR's source code is available on GitHub at scverse/anndataR under the MIT license. It is compatible with R version 4.5, has been archived at 10.5281/zenodo.18775712 and included within Bioconductor: 10.18129/B9.bioc.anndataR. Installation instructions and tutorials can be found in the online documentation at anndatar.scverse.org. Issues can be reported at the GitHub repository. Code to reproduce the analyses performed can be found on GitHub at LouiseDck/anndataR-paper, archived at 10.5281/zenodo.18792241.
IDAP: an integrated literature- and knowledge-graph-driven evidence prioritization pipeline for precision oncology.
Ryu Y, Jeong HE, An JY
Bioinformatics · 2026 May · PMID 42105215 · Full text
MOTIVATION: Advances in tumor sequencing enable routine detection of dozens to hundreds of somatic alterations per patient, yet only a minority can be linked to established therapeutic evidence. Curated resources such as...
MOTIVATION: Advances in tumor sequencing enable routine detection of dozens to hundreds of somatic alterations per patient, yet only a minority can be linked to established therapeutic evidence. Curated resources such as OncoKB provide high-quality variant-drug annotations but remain limited in coverage, particularly for rare or low-frequency variants. This coverage gap motivates computational frameworks that can integrate curated, literature-derived, graph-based, and clinical-trial evidence to prioritize therapeutic hypotheses for expert review. RESULTS: We developed the Integrated Drug Annotation Pipeline (IDAP), a modular framework that combines four complementary evidence streams: curated variant-drug associations from OncoKB, literature-derived gene-drug mention counts from PubMed abstracts, graph-based drug prioritization using a TxGNN-derived biomedical knowledge graph, and cancer-specific clinical-trial evidence from ClinicalTrials.gov. Given a cancer type and a MAF file, IDAP generates patient-level reports summarizing detected variants, ranked therapeutic hypotheses, supporting evidence layers, and relevant clinical trials. Evaluated across five cancer types (n = 50 samples), IDAP expanded evidence-linked therapeutic hypotheses beyond curated databases alone. Among patients without OncoKB recommendations (26/50), IDAP identified a median of 87 candidate drugs (range: 2-473). To reduce cross-source scale imbalance, the final ranking used within-sample percentile normalization with fixed bonuses for curated evidence, multi-source support, and trial linkage. Under this revised ranking, 24/50 top-ranked candidates were supported by at least two evidence sources and 44/50 had associated ClinicalTrials metadata. In an exploratory external CIViC comparison, IDAP recovered at least one matched CIViC-supported therapy in 28/41 eligible samples, with 13/41 appearing within the top 10 candidates. These outputs are intended to support evidence triage and translational interpretation rather than direct treatment recommendation. AVAILABILITY AND IMPLEMENTATION: IDAP is freely available at https://github.com/joonan-lab/IDAP-pipeline, with full documentation at https://joonan-lab.github.io/IDAP-pipeline. An archived snapshot of the code used in this study is deposited on Zenodo (DOI: https://doi.org/10.5281/zenodo.19301367).
TSEDTA: a transformer-based neural network with SMILES transformer and ESM2 embeddings for drug-target binding affinity prediction.
Sun X, Liu X, Huang J … +3 more , Wu J, Sun Y, Jia J
Bioinformatics · 2026 May · PMID 42105214 · Full text
MOTIVATION: Drug-target binding affinity (DTA) prediction plays a vital role in drug repositioning. The emergence of large language models (LLMs) has introduced new perspectives for predicting DTA. Herein, we present TSE...
MOTIVATION: Drug-target binding affinity (DTA) prediction plays a vital role in drug repositioning. The emergence of large language models (LLMs) has introduced new perspectives for predicting DTA. Herein, we present TSEDTA, a Transformer-based neural network with SMILES Transformer and ESM2 embeddings for predicting DTA. It leverages pre-trained LLMs (SMILES Transformer and ESM2) to extract deep evolutionary representations from drug SMILES and protein sequences. The representations are directly fused with raw sequence embeddings and processed via dual Transformer encoders to capture complex local and global dependencies. RESULTS: The experiments demonstrate that TSEDTA outperforms ten advanced models on the Davis and KIBA datasets, and seven on the BindingDB dataset. Ablation studies show that incorporating LLM embeddings significantly improves the performance of TSEDTA. Furthermore, a practical case study demonstrates its real-world applicability. Ultimately, TSEDTA provides a highly accurate, robust tool for DTA prediction, offering new insights into the application of LLMs for DTA tasks. AVAILABILITY: The source code and data are available at: https://github.com/SunXu24Math/TSEDTA. The version of record is archived in Zenodo with the DOI: 10.5281/zenodo.19103249.
Protein language model embeddings improve HIV drug resistance prediction: a comprehensive benchmark with attention-based interpretability.
Farquhar H
Bioinformatics · 2026 May · PMID 42105210 · Full text
MOTIVATION: Accurate prediction of HIV drug resistance from viral sequences is critical for optimizing antiretroviral therapy. Traditional machine-learning approaches using binary mutation encoding achieve strong accurac...
MOTIVATION: Accurate prediction of HIV drug resistance from viral sequences is critical for optimizing antiretroviral therapy. Traditional machine-learning approaches using binary mutation encoding achieve strong accuracy but may fail to capture epistatic interactions and structural features relevant to resistance mechanisms. Protein language models (PLMs) offer learned representations encoding evolutionary and structural information, but have not been systematically benchmarked for HIV resistance prediction across the modern antiretroviral drug set. RESULTS: We evaluated ESM-2 (650 M parameters) with attention-weighted pooling for predicting resistance to 18 drugs across three classes (protease inhibitors, NRTIs, NNRTIs) on the Stanford HIVDB dataset (n = 6308 sequences). Attention-weighted ESM-2 embeddings significantly outperformed XGBoost baselines with binary mutation encoding (mean AUC 0.968 versus 0.955, P = .0017), with gains across 15 of 18 drugs and the largest improvements for drugs with complex resistance patterns. Attention weights showed 2.48-fold enrichment at known drug-resistance-mutation positions (P < .05 for 63% of drugs; NRTIs strongest at 4.20-fold). External validation on a 20% holdout showed minimal degradation (AUC 0.934). Benchmarking against ESM C 600M and ESM-1v confirmed performance is robust to PLM choice (mean AUC 0.942-0.946 across backbones). Performance was maintained across HIV-1 subtypes (B 0.924; B-divergent 0.900; non-B 0.884) and a temporal holdout (AUC 0.930). AVAILABILITY AND IMPLEMENTATION: Source code is available at https://github.com/hayden-farquhar/HIV-ESM-2 under an MIT license and archived at https://doi.org/10.5281/zenodo.19466629. Stanford HIVDB genotype-phenotype data are publicly available at https://hivdb.stanford.edu/.
dAMN: a genome-scale neural-mechanistic hybrid model to predict bacterial growth dynamics.
Faulon JL, Dursoniah D, Ahavi P … +2 more , Raynal A, Asin-Garcia E
Bioinformatics · 2026 May · PMID 42104120 · Full text
SUMMARY: This study presents dAMN, a genome-scale neural-mechanistic hybrid model that combines neural networks with dynamic flux balance analysis to predict bacterial growth dynamics across diverse nutrient environments...
SUMMARY: This study presents dAMN, a genome-scale neural-mechanistic hybrid model that combines neural networks with dynamic flux balance analysis to predict bacterial growth dynamics across diverse nutrient environments. Using a residual network architecture, dAMN predicts reaction fluxes and lag-phase parameters from initial medium composition, then integrates these predictions under stoichiometric constraints derived from genome-scale metabolic models. Trained on Escherichia coli and Pseudomonas putida growth datasets across combinatorial media, dAMN accurately forecasts temporal growth dynamics and generalizes to unseen media conditions, with mean R² ≥ 0.9. The model also reproduces biologically relevant behaviors including substrate depletion, acetate overflow, and diauxic shifts, while explicitly modeling lag phases usually absent from standard dFBA. AVAILABILITY AND IMPLEMENTATION: The dAMN software, associated models, and datasets are available at https://github.com/brsynth/dAMN-main-release and via Zenodo DOI: 10.5281/zenodo.17908125.
Utilization of long-read sequencing for the detection of structural rearrangements with AgileStructure.
Lascelles C, Raynor M, Crinnion LA … +5 more , Rose AMS, Diggle CP, Poulter JA, Watson CM, Carr IM
Bioinformatics · 2026 May · PMID 42104046 · Full text
MOTIVATION: Changes in genome organisation contribute to genetic disease when they disrupt gene function or regulation. Structural rearrangements may interrupt coding sequence or alter expression through promoter loss or...
MOTIVATION: Changes in genome organisation contribute to genetic disease when they disrupt gene function or regulation. Structural rearrangements may interrupt coding sequence or alter expression through promoter loss or gain, chromatin changes, copy-number variation, or disruption of short-range regulatory elements. Although short-read sequencing excels at detecting small variants, it performs poorly at resolving breakpoints of large rearrangements, especially in repetitive or low-complexity regions. Long-read sequencing overcomes these limitations, but analytical tools have not kept pace, making accurate identification and annotation of large structural variants challenging. RESULTS: We developed AgileStructure, a desktop application for locating and annotating large‑scale genomic rearrangements using aligned long‑read data. The software enables user‑guided exploration of breakpoint‑spanning reads, supporting accurate interpretation of complex events and filling a key gap in current structural variant analysis workflows. AVAILABILITY AND IMPLEMENTATION: Source code, binaries, user guide, and example aligned read data, are available on GitHub: https://github.com/msjimc/AgileStructure. An archived version is also available on Zenodo at https://doi.org/10.5281/zenodo.18610110.
Drug response profile-based machine learning enables strategic cell line and compound selection for drug development.
Abdel-Rehim A, Tate E, Soldatova LN … +1 more , King RD
Bioinformatics · 2026 Jun · PMID 42103986 · Full text
MOTIVATION: Early-stage drug discovery relies on testing compounds across a limited set of cell lines, making it challenging to capture biological diversity while maintaining experimental efficiency. Current predictive a...
MOTIVATION: Early-stage drug discovery relies on testing compounds across a limited set of cell lines, making it challenging to capture biological diversity while maintaining experimental efficiency. Current predictive approaches for identifying responsive cell lines often depend on high-dimensional omics data, which can be costly and difficult to interpret. We therefore evaluated whether drug-response panel (DRP) descriptors, which capture sensitivity profiles to a reference set of compounds, can provide an efficient and informative alternative for modelling drug response in cell lines. RESULTS: Using gradient boosting models across GDSC and CCLE datasets, DRP descriptors consistently outperformed mRNA expression features in predicting drug sensitivity (-log10(IC50)), although performance varied across compounds. DRP-guided cell line selection enabled downstream omics-based modelling that recovered known MAPK-associated sensitivity signatures and identified potential biomarkers for MEK1/2 and BTK/MNK inhibitors. Extending this framework, we demonstrated its utility in compound prioritisation by distinguishing between tumourigenic MCF7 and non-tumourigenic MCF10A cells, successfully identifying compounds with selective activity. Together, these results show that DRP-based representations, derived from compact screening panels, support efficient cell line selection, biomarker discovery, and compound prioritisation in early-stage drug development. AVAILABILITY: Code and data uploaded to https://github.com/abbiAR/-Strategic-Cell-Line-and-Compound-Selection-Using-Drug-Response-Profiles.
KG-bench: benchmarking graph neural network algorithms for drug repurposing.
Wei S, Sasi C, Piepenbrock J … +3 more , Huynen MA, 't Hoen PAC, SIMPATHIC Consortium
Bioinformatics · 2026 May · PMID 42103971 · Full text
MOTIVATION: Drug repurposing leverages existing drugs for new indications, accelerating drug development. Computational methods integrating diverse biological and chemical data can systematically prioritize repurposing c...
MOTIVATION: Drug repurposing leverages existing drugs for new indications, accelerating drug development. Computational methods integrating diverse biological and chemical data can systematically prioritize repurposing candidates, but standardized benchmarks for deep learning evaluation are lacking. We present knowledge graph (KG)-Bench, a graph neural network (GNN) benchmarking framework designed to systematically compare the performance of different GNN architectures on drug-disease association prediction using the Open Targets dataset. We constructed a KG of drugs, diseases, and targets, including annotations such as therapeutic area and molecular pathway, and ensured retrospective validation by leveraging regular dataset updates. To avoid data leakage, we removed redundant entities across splits. RESULTS: Benchmarking six GNN architectures, Relational Graph Convolutional Networks achieved the highest ranking performance (AUC: 0.91), while TransformerConv showed superior robustness under class imbalance (F1: 0.28 at 1:100 positive: negative ratio), characteristic of real drug repurposing datasets. KG-Bench also assesses bias, node/feature importance, and uses GNNExplainer for interpretability. Our open-source framework enables fair, reproducible evaluation of graph-based drug repurposing algorithms. AVAILABILITY AND IMPLEMENTATION: Data and codes are available at https://github.com/cmbi/Benchmark_GNN_OpenTargets.
EMTscore infers divergent EMT pathways from omics data and enables rapid screening for EMT-associated gene sets.
Wen H, Bleris L, Hong T
Bioinformatics · 2026 May · PMID 42097304 · Full text
MOTIVATION: Quantitative analyses of epithelial-mesenchymal transition (EMT) have been widely used in several areas of biomedical sciences due to its importance in development and cancer progression, but its multi-contex...
MOTIVATION: Quantitative analyses of epithelial-mesenchymal transition (EMT) have been widely used in several areas of biomedical sciences due to its importance in development and cancer progression, but its multi-contextual nature requires standardization and implementation of gene set scoring methods beyond capacities of conventional tools. RESULTS: We developed EMTscore, a package that provides an efficient implementation of unbiased scoring methods for multiple EMT pathways using individual single-cell or bulk omics data, and the package allows rapid screening for cellular processes correlated with EMT. AVAILABILITY AND IMPLEMENTATION: EMTscore is available from GitHub https://github.com/wenmm/EMTscore under the GNU General Public License, and is uploaded on Zenodo with a DOI 10.5281/zenodo.19487376.
StrainMake: reproducible hybrid metagenomics with MAG recovery and strain-level resolution.
Hennecart B, Belda E, de Lahondès R … +2 more , Zucker JD, Prifti E
Bioinformatics · 2026 May · PMID 42097292 · Full text
SUMMARY: Metagenomic workflows involve complex multi-step analyses, from quality control and assembly to binning, annotation, and strain-level profiling. Few existing metagenomic pipelines achieve the combination of flex...
SUMMARY: Metagenomic workflows involve complex multi-step analyses, from quality control and assembly to binning, annotation, and strain-level profiling. Few existing metagenomic pipelines achieve the combination of flexibility, reproducibility, and hybrid assembly support within a unified workflow. We present StrainMake, a Snakemake-based workflow for de novo metagenomic analysis from short, long, or hybrid sequencing data. StrainMake integrates widely used tools across all major steps-quality control, assembly, binning, dereplication, taxonomic and functional annotation-while also providing non-redundant gene catalogues, community-scale metabolic models, and strain-level microdiversity metrics. The modular design enables the use of alternative tools, scalable execution on HPC systems, and full reproducibility through Snakemake and Conda. RESULTS: Applied to the CAMI II strain-madness dataset, StrainMake produced high-quality assemblies and metagenome-assembled genomes (MAGs), while enabling strain-resolved comparisons across samples. Hybrid assemblies improved contiguity, whereas short-read assemblies offered faster runtimes, illustrating the workflow's benchmarking capacity. AVAILABILITY AND IMPLEMENTATION: StrainMake is open source and available at https://github.com/UMMISCO/strainmake, together with comprehensive documentation. Generated data are deposited in Zenodo (doi: 10.5281/zenodo.16950162).
Integrating plant phenotypic and genotypic data in the AGENT project: a BrAPI service implementation.
Brouwer M, Bauernfeind J, Davuluri G … +9 more , García Brizuela J, König P, Kumar S, Lange M, Weise S, Wijnker E, Pommier C, Ruff J, Kersey PJ
Bioinformatics · 2026 May · PMID 42097287 · Full text
MOTIVATION: The AGENT project established a network of actively cooperating European genebanks, integrating genomic and phenotypic data from accessions of wheat and barley. Due to specific storage demands for phenotypic...
MOTIVATION: The AGENT project established a network of actively cooperating European genebanks, integrating genomic and phenotypic data from accessions of wheat and barley. Due to specific storage demands for phenotypic and genotypic data, the project used separate database instances and backend technologies to manage integrated phenotypic and genotypic data. RESULTS: We discuss the challenges encountered when integrating dispersed data to serve through a single interface such as the Plant Breeding Application Programming Interface, BrAPI. We examine how the consistent mappability of genebank data to the BrAPI model can enable the implementation of effective services. The advantages of BrAPI in transparently linking distributed data entities through embedded, unique identifiers are highlighted. We present a technical solution involving a BrAPI proxy, which combines and merges separate BrAPI endpoints. Finally, we demonstrate the AGENT BrAPI implementation with an illustrative example that validates a suggested SNP for a trait from the literature by linking phenotypic, genotypic and passport data. AVAILABILITY AND IMPLEMENTATION: The BrAPI proxy implementation and documentation is available at the Python Package Index (https://pypi.org/project/brapi-proxy) and archived in Zenodo (doi: 10.5281/zenodo.19436445). SUPPLEMENTARY INFORMATION: A Jupyter Notebook file for the validation example using a marker-trait relationship found in the literature.
PyFgsea: a Rust-powered, fgseaMultilevel-aligned GSEA framework with rolling-window enrichment along single-cell trajectories.
Wang K, Shi H
Bioinformatics · 2026 May · PMID 42095604 · Full text
SUMMARY: GSEA is a standard approach for pathway interpretation, yet Python ecosystems lack a high-performance implementation aligned with the fgseaMultilevel rare-event estimator target, especially for trajectory-aware...
SUMMARY: GSEA is a standard approach for pathway interpretation, yet Python ecosystems lack a high-performance implementation aligned with the fgseaMultilevel rare-event estimator target, especially for trajectory-aware rolling-window analysis. Under matched inputs, PyFgsea remains near-identical for normalized enrichment scores (NES; Pearson r>0.999), machine-precision identical for enrichment scores (ES), and statistically faithful for nominal P values relative to the R fgseaMultilevel reference. Its stateful rolling-window engine further reduces repeated trajectory-analysis overhead, yielding ∼1.9-fold end-to-end wall-time speedup in a conservative stress test and, in a narrower 100-window component benchmark, up to 7.47-fold acceleration. Rolling-window significance is controlled only by within-window Benjamini-Hochberg correction across pathways rather than by trajectory-wide global error control, so these profiles are intended primarily for local trend exploration and candidate-pathway prioritization. AVAILABILITY AND IMPLEMENTATION: Source code is available at https://github.com/shayuanxukuang/pyfgsea and via PyPI (pip install pyfgsea). An archival snapshot of the code and benchmark data is available on Zenodo (DOI: 10.5281/zenodo.19446446).
Disentangled diffusion model for 3D molecular generation with protein-ligand interaction priors.
Huang Z, Yang L, Qin C … +7 more , Xing Y, Yu H, Zhou X, Zheng B, Wang Y, Gao X, Yang W
Bioinformatics · 2026 Jun · PMID 42087329 · Full text
MOTIVATION: Structure-based drug design (SBDD) aims to generate ligand molecules that tightly bind to specific protein targets, a critical step in drug discovery. Diffusion models have shown promise for this task, yet ex...
MOTIVATION: Structure-based drug design (SBDD) aims to generate ligand molecules that tightly bind to specific protein targets, a critical step in drug discovery. Diffusion models have shown promise for this task, yet existing methods struggle to effectively incorporate protein-ligand interaction priors during generation. Most approaches rely on protein-specific structural priors that remain fixed throughout generation, limiting molecular diversity and failing to capture the dynamic interplay between protein pockets and ligand atoms, which is essential for achieving high binding affinity. RESULTS: We propose DPDiff, a disentangled prior-conditioned diffusion model for protein-specific 3D molecular generation. DPDiff introduces two complementary interaction prior networks that capture geometry-based spatial interactions and sequence-based interactions robust to structural noise. During generation, the model dynamically extracts interaction priors using intermediate diffusion predictions and adaptively fuses them via a time-dependent adapter. A disentangled denoising network balances prior guidance with generative flexibility. Experiments on the CrossDocked2020 dataset demonstrate that DPDiff generates molecules with more realistic 3D structures and state-of-the-art binding affinities, achieving an average Vina Dock score of -8.58 and a high affinity ratio of 69.4%, outperforming existing methods while maintaining favorable drug-likeness and synthetic accessibility. AVAILABILITY AND IMPLEMENTATION: The source code of DPDiff is available at https://github.com/ZerinHwang03/DPDiff.
Segzoo: a turnkey system that summarizes genome annotations.
Mendez M, Liu Y, Asenjo Ponce de León M … +1 more , Hoffman MM
Bioinformatics · 2026 May · PMID 42087325 · Full text
MOTIVATION: Segmentation and automated genome annotation (SAGA) techniques, such as Segway and ChromHMM, assign labels to every part of the genome, identifying similar patterns across multiple genomic input signals. Infe...
MOTIVATION: Segmentation and automated genome annotation (SAGA) techniques, such as Segway and ChromHMM, assign labels to every part of the genome, identifying similar patterns across multiple genomic input signals. Inferring biological meaning in these patterns remains challenging. Doing so requires a time-consuming process of manually downloading reference data, running multiple analysis methods, and interpreting many individual results. RESULTS: To simplify these tasks, we developed the turnkey system Segzoo. As input, Segzoo only requires a genome annotation file in browser extensible data (BED) format. It automatically downloads the rest of the data required for comparisons. Segzoo performs analyses using these data and summarizes results in a single visualization. AVAILABILITY AND IMPLEMENTATION: The source code for Python ≥ 3.7 on Linux is freely available for download at https://github.com/hoffmangroup/segzoo under the GNU General Public License (GPL) version 2. Segzoo is also available in the Bioconda package segzoo: https://anaconda.org/bioconda/segzoo. We have deposited in Zenodo the version of the Segzoo source which produced the results in this article (https://doi.org/10.5281/zenodo.10988775), other code and data used to produce the results (https://doi.org/10.5281/zenodo.10477083), and the results (https://doi.org/10.5281/zenodo.10477106).
Learning protein representations with conformational dynamics.
Kalifa D, Horvitz E, Radinsky K
Bioinformatics · 2026 May · PMID 42085496 · Full text
MOTIVATION: Proteins change shape as they work, and these changing states control whether binding sites are exposed, signals are relayed, and catalysis proceeds. Most protein language models (PLMs) pair a sequence with a...
MOTIVATION: Proteins change shape as they work, and these changing states control whether binding sites are exposed, signals are relayed, and catalysis proceeds. Most protein language models (PLMs) pair a sequence with a single structural snapshot, which can miss state-dependent features central to interaction, localization, and enzyme activity. Studies also indicate that many proteins assume multiple, functionally relevant shapes, motivating approaches that learn from this variability. RESULTS: We present DynamicsPLM, a PLM conditioned on ensembles of computationally generated conformations to derive state-aware representations. DynamicsPLM improves predictive performance across protein-protein interaction, subcellular localization, enzyme classification, and metal-ion binding. On a widely used protein-protein interaction benchmark, it achieves a four-point accuracy gain over the strongest baseline. On a curated test set enriched for proteins with multiple conformational states, the margin increases to eleven points. These findings argue for a shift from static to dynamics-aware modeling, in which conformational variability is treated as informative. By elevating conformational state to a central element of machine learning in protein biology, this work advances modeling toward mechanisms that better reflect how proteins operate in cells and provides a route to actionable hypotheses about when and how binding, signaling, and catalysis occur. AVAILABILITY AND IMPLEMENTATION: Code, model weights, and inference scripts are available at https://github.com/kalifadan/DynamicsPLM (DOI: https://doi.org/10.5281/zenodo.17668302).
FuFiHLA: a tool for full-field HLA typing from long-read data.
Hu J, Qin Q, Li H … +1 more , Zhou Y
Bioinformatics · 2026 May · PMID 42085485 · Full text
MOTIVATION: Allele typing for Human Leukocyte Antigen (HLA) genes has many important clinical applications. Popular short-read typing can only accurately distinguish alleles at the coding sequence level, which potentiall...
MOTIVATION: Allele typing for Human Leukocyte Antigen (HLA) genes has many important clinical applications. Popular short-read typing can only accurately distinguish alleles at the coding sequence level, which potentially limit our understanding of the effect of variants in non-coding region. Long read data has been proved to be useful in typing HLA alleles in full resolution, but only a few tools are publicly available and with significant limitations in practical application. RESULTS: We developed FuFiHLA, a lightweight open-source software, to type HLA alleles. Currently it supports typing alleles of six HLA genes (HLA-A, HLA-B, HLA-C, HLA-DRB1, HLA-DQA1, and HLA-DQB1) from long reads. Evaluation using 233 PacBio HiFi WGS samples from HPRC shows that FuFiHLA achieves 99.6% accuracy in the full field allele typing and QV as 51.8 for consensus allele sequence construction. Additional testing on four Nanopore R10 reads demonstrates slightly reduced accuracy in the fourth field. AVAILABILITY: FuFiHLA is available at https://github.com/jingqing-hu/FuFiHLA under MIT License.
Learning drug synergy through environment-conditioned feature modulation.
Jin S, Huang A, Meng Y … +4 more , Zhu Z, Jiang Y, Xu J, Zeng X
Bioinformatics · 2026 May · PMID 42085481 · Full text
MOTIVATION: Drug combinations are crucial for overcoming resistance in cancer therapy. Although deep learning has achieved strong performance in synergy prediction, existing models often treat cell-specific features and...
MOTIVATION: Drug combinations are crucial for overcoming resistance in cancer therapy. Although deep learning has achieved strong performance in synergy prediction, existing models often treat cell-specific features and paired drugs as a static background and fail to capture how the specific cell-drug environment dynamically modulates drug representations, thereby hindering the modeling of environment-specific synergistic effects. RESULTS: We propose Env-Syn, a framework for modeling drug-drug-cell interactions through Environment-Conditioned Feature Modulation, which incorporates a Residual Feature-wise Linear Modulation (R-FiLM) module to perform precise affine transformations on drug representations conditioned on paired drugs and cellular environments. Benchmark evaluations show that Env-Syn consistently outperforms state-of-the-art methods. Notably, the model exhibits exceptional generalization performance in rigorous inductive scenarios. It maintains high predictive accuracy for unseen drugs with AUROC and AUPRC exceeding 0.81 in the Leave-drug-out setting and further demonstrates strong cross-dataset reliability by surpassing a recall of 0.7 on independent test set. Furthermore, among 15 novel predicted drug combinations, 8 are directly supported by literature evidence. These results demonstrate that Env-Syn is an effective computational tool for drug synergy discovery. AVAILABILITY AND IMPLEMENTATION: The source code is available at https://github.com/AnQi-87/Env-Syn.
FourierDrug: a domain generalization framework for robust drug response prediction via frequency-space asymmetric attention.
Song R, Bai Y, Liu X … +1 more , Liu H
Bioinformatics · 2026 Jun · PMID 42085479 · Full text
MOTIVATION: Accurate prediction of drug response remains a major challenge in precision oncology, particularly at the single-cell level and in clinical settings, due to significant distribution shifts between preclinical...
MOTIVATION: Accurate prediction of drug response remains a major challenge in precision oncology, particularly at the single-cell level and in clinical settings, due to significant distribution shifts between preclinical models and real-world patient data. Existing approaches often rely on transfer learning from cell lines to target domains, but typically require access to target-domain data during training, which is frequently unavailable in practice. RESULTS: We propose FourierDrug, a novel domain generalization framework for robust drug response prediction. Given gene expression profiles, the model performs Fourier transformation to project features into the frequency domain and introduces an asymmetric attention mechanism that encourages drug-sensitive samples to form compact clusters while driving resistant samples to be more dispersed. This design facilitates the learning of domain-invariant yet task-relevant representations. Extensive experiments demonstrate that FourierDrug effectively leverages diverse source domains and generalizes well to unseen cancer types. Notably, when evaluated on single-cell and patient-level prediction tasks, our method-trained solely on in vitro cell line data without access to target-domain data-consistently outperforms or matches state-of-the-art approaches. AVAILABILITY AND IMPLEMENTATION: The source code and processed datasets are available at: https://github.com/hliulab/FourierDrug.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe