Latest bioRxiv papers

Category: bioinformatics — Showing 50 items

• A Systematic Comparison of Single-Cell Perturbation Response Prediction Models

  Li, L.; You, Y.; Fu, Y.; Liao, W.; Fan, X.; Lu, S.; Cao, Y.; Li, B.; Ren, W.; Kong, J.; Zheng, S.; Chen, J.; Liu, X.; Tian, L.

  Abstract

  Predicting single-cell transcriptional responses to perturbations is central to dissecting gene regulation and accelerating therapeutic design, yet the field lacks a rigorous, task-spanning assessment of model behavior. We present a large-scale benchmark of 13 representative methods and baselines across 25 datasets spanning diverse perturbation modalities and species, including two primary immune-cell drug-response resources. We evaluated three core tasks-generalization to unseen single-gene perturbations, prediction of combinatorial interactions, and transfer across cell types-using 24 metrics covering expression-level accuracy, relative changes, differential expression recovery, and distributional similarity.Across tasks, performance depended strongly on perturbation effect size and evaluation perspective: expression-level agreement was highest for small-effect perturbations resembling controls, whereas delta- and DE-based metrics improved with larger effects, providing clearer signals. Models shared a conservative bias, with fine-tuned foundation models compressing variance and underestimating synergistic effects in combinations. PerturbNet showed superior recovery of DE signatures in Tasks 1 and 2, while no method consistently generalized across cell types in Task 3, where biological consistency dominated outcomes.This benchmark establishes current methodological limits, clarifies that different metrics probe distinct biological signals rather than redundant summaries of the same prediction problem, and provides a foundation for developing virtual-cell models that more faithfully capture heterogeneous perturbation responses.

  bioinformatics2026-08-03v3

• Spatial Transcriptomics of Glioblastoma Defines Spatially Anchored Transcriptional Activity Informing Therapeutic Vulnerability and Resistance

  Ravikumar, V.; Maddox, A.; Kulkarni, R.; Hara, T.; Rao, A.; Al-Holou, W.

  Abstract

  Intratumoral heterogeneity is thought to hinder targeted therapy in glioblastoma (GBM), but the extent of variation in known targets and tumor cell-intrinsic and extrinsic regulatory mechanisms within tissue remains unclear. In this work, we use Visium spatial transcriptomics to profile 44 tumor slides and 128,176 niche-annotated ST spots, including 14 new slides from 7 GBM specimens, along with spatial annotation from the Ivy Glioblastoma Atlas Project. First, we characterize the differential expression of AVB3, EGFR, VEGF, and PDGFRA pathway activities across space, elucidating the limited success of individual targeted therapies. Second, we identify transcription factor modules driving intratumoral heterogeneity and characterize a novel transcription factor module at the interface between core tumor and the invasive edge. Third, we identify cell type-specific receptor-ligand signaling enriched in specific spatial niches and relate these to specific molecular program alterations. Finally, we identify and characterize transcriptional pathways and their activity across spatial niches through transcription network analysis and demonstrate how these pathways can inform a combined therapeutic approach. Together, through novel insights into the spatial patterning of transcription factor regulation, cellular interactions, and biological pathway activity, our work informs rational combination therapies targeting spatial niche specific vulnerabilities.

  bioinformatics2026-08-03v3

• BioGraphX-RNA: A Universal Physicochemical Graph Encoding for Interpretable RNA Subcellular Localization Prediction

  Saeed, A.; Abbas, W.

  Abstract

  RNA subcellular localization is a critical determinant of cellular function. However, current computational approaches often operate as ""black boxes", overlooking the complex interplay among sequence, structure, and physicochemical interactions that govern RNA localization. Building upon the BioGraphX framework originally developed for proteins, we introduce BioGraphX-RNA, a universal physicochemical graph-encoding framework that provides a structure-informed encoding by translating primary nucleotide sequences into multiscale interaction graphs using explicit biophysical rules. When combined with frozen RiNALMo embeddings via an interpretable gated fusion layer, BioGraphX-RNA achieves competitive performance with DeepLocRNA and uniquely quantifies the relative contribution of sequence versus structure for each RNA. On human datasets, the gated fusion model attains macro AUROC values of 0.7575 +/-0.0054 (mRNA), 0.9228 +/- 0.0137(miRNA), and 0.5600 +/- 0.0191 (lncRNA). For miRNA, the graph only model alone reaches 0.9396 +/-0.0045, outperforming both the RiNALMo language model and a RNAfold partition function graph (0.9139 +/- 0.0138), validating the structure;informed proxy hypothesis. In a blind cross-species prediction task on mouse data, the model shows limited zeroo shot transfer, indicating that biophysical graph features do not improve cross species generalization. Gating analysis reveals RNA type-specific modality reliance, with miRNA exhibiting a near equilibrium balance between sequence and structure. SHAP based interpretation suggests potential correlates such as patterned GC content for nuclear retention and structural accessibility for exosome targeting. These advances are achieved with only 2.05 million trainable parameters, aligning with Green AI principles. BioGraphX-RNA demonstrates that explicitly integrating biophysical constraints into graph-based encodings enables accurate and interpretable predictions for structured RNAs, advancing structure-aware RNA biology and laying a foundation for precision medicine.

  bioinformatics2026-08-03v3

• RKMR: A Rapid Kernel Machine Regression Framework for Optimal Marker Detection in Spatial Omics Data

  Seal, S.; Chakraborty, A.; Mattila, C.; Rubinstein, M.; Angel, P.; Ghosh, D.; Chung, D.; Neelon, B.

  Abstract

  High-throughput spatial omics technologies enable molecular profiling within intact tissue architecture, yet identifying concise, predictive, and biologically interpretable marker panels for cell types, tissue domains, and disease-associated tissue classes remains challenging. This limitation hinders the development of actionable panels for targeted validation and downstream translation. Existing pipelines rely largely on univariate differential-expression analyses, which ignore joint molecular structure and provide limited predictive insight. Multivariate machine-learning methods, including random forest, XGBoost, elastic net, and specialized single-cell panel-selection approaches, can capture predictive patterns but typically lack explicit spatial modeling and probabilistic feature selection, relying instead on model-specific importance scores or user-specified panel sizes. We develop rapid kernel machine regression (RKMR), a scalable framework for spatial-omics marker discovery that integrates nonlinear kernel modeling, spike-and-slab variable selection, and spatial dependence. RKMR uses automatic relevance determination (ARD) kernels and sparsity-inducing priors to capture nonlinear marker-outcome relationships and implicit feature interactions while producing approximate posterior inclusion probabilities (PIPs) that quantify model-based uncertainty in feature inclusion. To scale inference to large spatial datasets, RKMR combines low-rank kernel approximations with stochastic variational optimization. In simulations, RKMR consistently achieves higher AUPRC than competing methods across a range of molecular-signal and spatial-effect settings. Across spatial transcriptomics and scRNA-seq datasets, RKMR identifies parsimonious marker sets that recover reported cell-type signatures and reproducible tissue-layer markers. These results establish RKMR as a scalable and uncertainty-aware framework for translating high-dimensional spatial omics data into robust, experimentally actionable marker panels.

  bioinformatics2026-08-03v3

• crispAIPE: Probabilistic Modelling of Prime Editing Variant Correction Efficiency

  Ozden, F.; Lu, P.; Minary, P.

  Abstract

  Prime editing installs precise substitutions, insertions, and deletions without double-strand breaks or donor templates, but the efficiency of an individual pegRNA design is hard to predict in advance, and existing tools give only point estimates, leaving designers unable to judge which predictions to trust. We present crispAIPE, a transformer-based probabilistic framework that quantifies per pegRNA-target pair uncertainty by modelling the three competing prime-editing outcomes on the 2-simplex with a Dirichlet likelihood and pairing the posterior with split-conformal highest-density regions calibrated on a held-out fold, giving finite-sample coverage guarantees on the simplex without requiring the Dirichlet model itself to be well-calibrated. Trained on 92,423 PRIDICT Library-1 pegRNAs under mutation-level target-disjoint splitting, crispAIPE attains Spearman rho = 0.835, 0.843, and 0.693 on the edited, unedited, and indel fractions (Pearson r = 0.845, 0.855, 0.659), and its conformal regions match nominal coverage where region-construction baselines do not. pegRNA architecture and edit context, in particular GC content of the mutated reverse transcription template (RTT) and edit size for deletions, are associated with prediction uncertainty. Reusing the Library-1 calibration quantile unchanged on PRIDICT Library-2, the conformal region area generalises as an actionable filter for cross-cell-type design, identifying the pegRNAs whose Library-1 predictions transfer best while small-sample head fine-tuning further improves accuracy. Tool and trained models: https://github.com/furkanozdenn/pe-uncert.

  bioinformatics2026-08-03v2

• Explainable Generative AI Uncovers a Molecular Continuum in Medulloblastoma with Implications for Rare Cancer Subtyping and Treatment Equity

  Prol-Castelo, G.; Tejada-Lapuerta, A.; Urda-Garcia, B.; Nunez-Carpintero, I.; Garcia-Verellen, E.; Montagud, A.; Valencia, A.; Cirillo, D.

  Abstract

  Medulloblastoma is a childhood brain tumor traditionally classified into four molecular subgroups. Recent evidence suggests that Groups 3 and 4 represent a biological continuum rather than distinct entities, a paradigm shift with significant implications for understanding disease biology and treatment strategies. Nevertheless, assessing this hypothesis is challenging mainly due to data scarcity. In this study, we analyze the largest available transcriptomics dataset to provide compelling evidence for the existence of an intermediate subgroup between Groups 3 and 4, characterized by distinct molecular features. To overcome limitations posed by data scarcity, we employ synthetic data generation using a Variational Autoencoder and apply explainability techniques to identify key relationships between gene expression and disease subgroups. Furthermore, by incorporating Machine Learning Fairness approaches, we demonstrate that overlooking this intermediate subgroup can result in treatment disparities. Our findings are further supported by both existing and newly proposed studies using diverse datasets and methodologies, including graph-based analyses and multi-scale simulations, underscoring the robustness and reproducibility of our results. This study demonstrates the potential of synthetic data generation to refine rare disease subtyping and advance our understanding of the underlying biological mechanisms.

  bioinformatics2026-08-03v2

• PG-LLM: Benchmarking General-Purpose Language Models for Protein Variant Ranking

  Arora, R. K.; Chen, L. T.; Du, M.; Marks, D.; Church, G.

  Abstract

  eneral-purpose language models are being increasingly utilized in protein-design workflows, yet their ability to evaluate variant effects remains unclear. To answer this question, we introduce PG-LLM, a benchmark built on ProteinGym to evaluate general-purpose language models on 217 protein-variant prioritization tasks. Each task follows the same format: a language model receives a wild-type protein sequence, an assay description, and is tasked with ranking 50 mutant sequences by fitness without access to multiple-sequence alignments or protein structures. We evaluate thirteen language models and rescore 95 published protein predictors on the same candidate sets with the same evaluation metric. Claude Opus 5 leads the primary leaderboard with a Spearman correlation of {rho} = 0.406, narrowly ahead of GPT-5.6 Sol at 0.402. However, GPT-5.6 Sol scores higher than Opus 5 when the two models are compared only on assays scored by both. Opus 5 outperforms 49 of 95 published protein predictors, including 41 of 46 sequence-only methods, and approaches ESM2-650M at {rho} = 0.411, but remains below the leading predictor VenusREM at {rho} = 0.523. Variant-ranking performance improves with test-time compute across GPT, Claude, and Gemini models, but the gains taper before closing the gap to specialist protein predictors. Unlike sequence-only predictors, which perform better on proteins with deeper evolutionary alignments, LLM accuracy changes little across alignment-depth. To address contamination risk, we additionally source 59 DMS assays from 19 studies whose scores first became public after January 2026. On this set, we evaluate three LLMs and seven biomolecular baselines, observing performance trends similar to those on the main PG-LLM benchmark. PG-LLM shows that tool-free language models capture substantial protein-variant signal and already outperform many established sequence-based predictors. These results establish the emerging capability of language models as biomolecular reasoners while defining the remaining headroom for their reliable use in variant-prioritization workflows.

  bioinformatics2026-08-03v2

• vartracker: an end-to-end tool for pathogen longitudinal variant analysis and visualisation

  Foster, C. S. P.; Rawlinson, W. D.

  Abstract

  Longitudinal sequencing can reveal fine-grained pathogen evolution during acute and chronic infections and inform public health responses. However, integrating ordered pathogen genomic data into a coherent evolutionary and clinical framework can be tedious and error-prone. We present vartracker, an open-source tool for longitudinal pathogen variant analysis and visualisation. Given an ordered sample manifest, vartracker supports three entry points: raw sequence reads, reference-aligned BAM files, or user-supplied VCF and coverage inputs. Raw-read and BAM inputs are processed through an integrated Snakemake workflow, whereas VCF mode starts from precomputed files. Variants are normalised and annotated relative to a reference genome, tracked across timepoints, and classified as original or newly emerging and as transient or persistent. Inferred amino acid changes are reported, and for SARS-CoV-2 analyses, relevant published literature for key mutations can be automatically linked through a functional database. vartracker outputs a schema-documented results table, provenance metadata for reproducibility, publication-quality static figures, and an interactive heatmap for data exploration. Although packaged with SARS-CoV-2 reference assets and initially developed for SARS-CoV-2 datasets, vartracker is pathogen-agnostic when appropriate reference data are supplied. We demonstrate its utility using SARS-CoV-2 and respiratory syncytial virus A (RSV-A) datasets. vartracker is freely available through GitHub, PyPI and Bioconda.

  bioinformatics2026-08-03v2

• Do Geometric Outliers Identify Important Genes in Single-Cell Foundation Models?

  Whalley, J. P.

  Abstract

  Single-cell foundation models (scFMs) produce gene embeddings that are increasingly interpreted biologically, but it is unclear whether different models organise gene space consistently, or whether geometric extremes identify important genes. The same four-metric screen was applied to Geneformer, scGPT and scFoundation, which differ in architecture, training objective and expression encoding. The models agreed weakly on individual outlier genes but showed structured class-level convergence: ribosomal enrichment occurred in all three, although weaker and caller-dependent in scFoundation, while strong mitochondrial enrichment was confined to Geneformer and scGPT. Gene-level agreement was itself uneven, with Geneformer and scGPT overlapping above chance and scGPT and scFoundation not, and caller stability differed between models. Most outliers were not extreme in ESM-2 protein-sequence space, indicating principally model-specific rather than protein-sequence geometry. These patterns did not translate into the tested notions of gene importance: deleting the highest-anomaly Geneformer genes did not impair cell-type annotation beyond matched controls, and covariate-adjusted outlier status was not associated with ClinVar membership in any model. Gene-embedding geometry therefore characterises model-specific structure, but does not provide standalone evidence of downstream leverage or disease relevance.

  bioinformatics2026-08-03v2

• What limits local ancestry inference at low divergence: a feasibility threshold, a metric that conceals failure, and a deficit of input more than architecture

  Tian, Q.

  Abstract

  Local ancestry inference assigns each position along an admixed chromosome to a source population, underpinning admixture mapping, ancestry-specific association testing and admixture dating. Validation is almost exclusively on continentally divergent sources (Hudson's FST {approx} 0.1) and coalescent simulations; we examine both restrictions. Across FST from 0.0022 to 0.243 we compare five methods--two likelihood baselines, RFMix, FLARE and a dilated convolutional network--on identical sites with exact ground truth, and on 11 real 1000 Genomes pairs. Three findings follow. First, a feasibility floor: at FST = 0.0022 no method exceeds 0.575, and CHB/CHS at FST = 0.00042 yields at best 0.551. Pairs motivating fine-scale analysis, such as northern versus southern Han, fall below it. Second, per-site accuracy conceals a failure of tract structure: the most accurate method per site produces 78.8x too many tracts, implying an admixture time 61.2x too old, which Viterbi decoding removes at no cost to accuracy (+0.0002). Third, the simulated lead does not survive real data, and the deficit is one of input more than of the architectures we varied: attention, state-space layers, capacity, objective and self-supervised pretraining each move accuracy by at most 0.006, while supplying the haplotype information the released tools receive recovers +0.031 on 8 of 8 pairs below FST = 0.04 and nothing above it--necessary but not sufficient, since the network still trails on 10 of 11 pairs. Two quantities usually held fixed matter more than architecture: the statistic summarising reference matching, and reference panel size, which no method is near saturating.

  bioinformatics2026-08-03v1

• STAR Suite: Transcriptomics processing in a single binary through AI-assisted development

  Hung, L.-H.; Baker, D.; Flynn, B.; Huangfu, D.; Luo, R.; Robson, P.; Zhou, T.; Yeung, K. Y.

  Abstract

  Single-cell transcriptomic assays - pooled CRISPR/Perturb-seq screens, clonal lineage tracing, and 10x Flex fixed RNA profiling - are central to modern biology, yet their processing has depended on Cell Ranger, a proprietary tool restricted to 10x products, with no open end-to-end pipeline for Perturb-seq or Flex. We present STAR Suite, which extends the STAR aligner into a single executable that processes bulk RNA-seq, scRNA-seq, Perturb-seq, 10x Flex, and SLAM-seq in one command, with no added dependencies. For single-cell assays it reproduces Cell Ranger 9.0.1 (gene-level Pearson 0.99-1.0) while running 3.6- to 5.7-fold faster, via novel exact-matching and alignment-validated hashing algorithms. Served to both AI agents and human operators through one validated interface, STAR Suite is the production standard for the NIH MorPhiC consortium; its 132,226 lines of C/C++ were built under a human-directed, AI-implemented workflow. Source code, workflow recipes, and per-run provenance are released under the MIT license

  bioinformatics2026-08-02v6

• MGMG: Cell Morphology-Guided Molecule Generation for Drug Discovery

  Tang, Q.; Ding, D.; Yuan, X.; Seabra, G.; Ramdhan, P. A.; Liu, C.-Y.; Thai, M. T.; Li, C.; Luesch, H.; Li, Y.

  Abstract

  Designing novel molecules with desired bioactivity remains a fundamental challenge in drug discovery. Most molecular design methods follow target-based drug discovery paradigms that rely on well-defined drug targets, thereby limiting their applicability to diseases lacking known targets or reference compounds. Here we introduce Morphology-Guided Molecule Generation (MGMG), a phenotypic drug discovery-oriented approach that integrates cellular morphological profiles from compound treatments with molecular textual descriptions without requiring molecular target information. Cell morphology offers the guidance on desired bioactivity-relevant phenotypic effects, while textual descriptions provide direct and interpretable cues about molecular structure. Leveraging complementary structural and bioactivity context, MGMG significantly enhances molecule generation performance, especially in scenarios where textual descriptions are under-informative or morphological signals are weak. MGMG can also be applied to genetic perturbations, enabling activator design from gene overexpression morphology without requiring knowledge of reference compound structure. In addition, in silico docking demonstrates that MGMG-generated molecules, despite lacking target information, exhibit binding affinities comparable to reference compounds, preserving key interactions while introducing structural diversity. Overall, MGMG jointly utilizes morphological and textual description inputs to guide molecule generation, enabling diverse, bioactivity-aware compound design in a target-agnostic fashion.

  bioinformatics2026-08-01v2

• Pangenome Graph Node-Phenotype Association shows GWAS-like quality results with only few individuals

  Carrette, C.; Sabot, F.; Muller, C.

  Abstract

  Purpose: We introduce GraNPA, standing for Graph Node-Phenotype Associa- tion, a method performing a GWAS-like analysis on a pangenome variation graph (PVG) built using a small number of individual genome sequences, without the need for additional population materials or kinship information for qualitative phenotypes. This method reduces the number of individuals required for associ- ation studies and prevents reference bias from variant calling in these types of analyses. Background: A PVG represents the multiple alignment of a set of complete genomes. It contains all variations, from single nucleotide polymorphisms (SNPs) to large structural variations (SVs), which are represented as nodes in the graph. By integrating phenotype information within nodes, we can assign a Phenotype Score (PS) to each node in the PVG and identify phenotype-related regions directly within it. These regions represent statistically significant shifts in PS distribution, highlighting their implication in the phenotype. Finally, GraNPA provides their positions and scores for further analysis. Results: This method was tested using simulated data and two publicly available datasets: the Sub1A gene locus for Oryza sativa in a 13 indi- viduals PVG, and the insertion responsible for the white-headed cattle with a PVG of 24 individuals. Source code of GraNPA is available here https://forge.ird.fr/diade/graphgwas/granpa under GNU GPLv3. Conclusion: GraNPA was able to identify the expected area in two simulated datasets and the responsible loci for these two known traits using only a few 1 dozen complete genomes in these PVGs. While currently limited to qualitative phenotypes, this method opens the way to more efficient ones relying on PVGs and few individuals.

  bioinformatics2026-08-01v1

• Cross-Attention Over RNA And Protein Sequences Enables Generalizable Interaction Prediction

  Catalano, M.; Pepe, G.; Appierdo, R.; Ausiello, G.; McWhite, C.; Gambosi, G.; Helmer-Citterich, M.; Gherardini, P. F.

  Abstract

  Computational predictions are essential to characterize the RNA-protein interaction landscape, yet a persistent gap between benchmark performance and practical utility suggests that current models have limited generalization capabilities. To address this issue, we present CORAL (Cross-attention for RNA-protein Association Learning), a deep learning framework for the prediction of RNA-protein interactions that integrates pretrained protein (ESM-2) and RNA (DNABERT2) language models through bidirectional cross-attention with Low-Rank Adaptation fine-tuning. We also introduce a benchmarking framework that rigorously addresses the problem of data redundancy between training and test sets, which greatly inflates model performances reported in the literature. To this end we adopt three partitioning strategies of increasing stringency: conventional random splits, pairwise non-redundant splits, and component-wise non-redundant splits. CORAL maintains an F1 score of 0.65 under the most stringent component-wise evaluation, compared to 0.47 for the next-best method retaining discriminative behavior. Interpretability analyses further reveal that the cross-attention mechanism captures biologically meaningful features of molecular recognition at two complementary structural scales: at atomic resolution, specific attention heads systematically attend to structurally defined contact positions, showing 26% elevated attention at interface residues across 309 experimentally resolved complexes (p < 0.001); at the domain level, protein-side attention localizes to annotated RNA-binding domains across 94% of 462 proteins examined (median within-protein Cohen's d {approx} 0.94). Together, these findings establish that current RPI prediction benchmarks substantially inflate performance estimates and demonstrate that cross-modal attention architectures yield improved generalization alongside mechanistically interpretable representations.

  bioinformatics2026-07-31v2

• DPCGS: a computational framework for linking GWAS to single-cell transcriptomics in complex traits and diseases

  Liu, C.; Yuan, B.; Shen, B.; Li, J.; Zhu, R.; Yang, P.; Wu, B.; Xuan, Y.; Yang, S.; Yang, N.; Ma, L.; Liu, Q.; Dai, S.; Zhang, Y.

  Abstract

  Complex traits and diseases arise from the interplay between genetic variation and cellular heterogeneity, making it essential to understand how genetic risk manifests at the cellular level. However, connecting genome-wide association studies (GWAS) to specific cell populations remains challenging due to cellular complexity and the prevalence of noncoding variants. Here, we present DPCGS, a computational framework that systematically integrates GWAS summary statistics with single-cell RNA-sequencing (scRNA-seq) data to identify trait-associated cell subpopulations, genes, and regulatory programs. Unlike existing approaches that primarily evaluate pathway enrichment or cell-type-level associations, DPCGS quantifies the enrichment of genetically prioritized genes within individual cells through a statistically calibrated gene-set scoring strategy, enabling high-resolution mapping of genetic risk to cellular states. Benchmarking across simulated and diverse human single-cell datasets demonstrates that DPCGS achieves superior accuracy, sensitivity, and robustness compared with existing methods, including scDRS and scPagwas. Applying DPCGS to Alzheimer's disease and asthma reveals disease-associated cellular populations and uncovers potential molecular drivers, including CD74, FOS, and AP-1 family regulatory programs, providing insights into disease-specific immune and cellular mechanisms. By bridging genetic discoveries from GWAS with functional interpretation at single-cell resolution, DPCGS establishes a generalizable framework for dissecting the cellular architecture of complex traits and diseases. This approach enables systematic discovery of disease-relevant cell subpopulations, regulatory networks, and potential therapeutic targets, offering broad applications in human genetics, single-cell biology, and precision medicine.

  bioinformatics2026-07-31v2

• Forecasting cell state futures from static snapshots

  Luo, E.; Gao, H.; BIAN, H.; Peng, S.; Li, Y.; Li, C.; Hao, M.; Chen, M.; She, Y.; Wei, L.; Liu, K.; Zhang, X.

  Abstract

  Predicting future cell state transitions from static transcriptomic snapshots is vital for understanding and steering cellular dynamics. Although existing trajectory and RNA velocity methods can infer local transitions within observed data, they are fundamentally limited in extrapolating long-range cellular dynamics. Here, we introduce CellTempo, an autoregressive generative framework that enables long-range forecasting of cell state trajectories directly from static transcriptomic snapshots. CellTempo is pretrained on scBaseTraj, a large-scale collection of multi-step cellular transition sequences constructed by integrating diverse experimental technologies and data sources. Across diverse biological systems, CellTempo accurately predicts long-range lineage progression and reconstructs unseen cellular potential landscapes. It further distinguishes transient perturbation responses from long-term cell-fate changes, enabling prioritization of candidate compounds for cell-fate engineering. These results demonstrate that long-range cellular dynamics are inferable from static observations, establishing a foundation for predictive and controllable modeling of cell fate.

  bioinformatics2026-07-31v2

• mock-fastq-generator: A synthetic FASTQ generator

  Florez Prada, A.; Hart, D. J.

  Abstract

  Validating bioinformatics pipelines and benchmarking sequence processing algorithms requires reliable test datasets. Existing read simulation tools rely on reference genomes and empirical error profiles, lacking fine-grained control over specific targeted DNA constructs and controlled error injection. mock-fastq-generator is an open-source software suite available both as an installable PyPI Python package and a standalone, client-side web application. It constructs synthetic FASTQ files by combining template constructs with customizable adapter sequences, selectable quality decay functions (Gaussian, Exponential, Sigmoidal), NovaSeq 4-state quality binning, and context-dependent error penalties. The software allows developers to benchmark sequence trimmers, test alignment sensitivity, and execute automated quality control pipelines in test suites without using proprietary or empirical biological data.

  bioinformatics2026-07-31v1

• MitoDate: a Nextflow pipeline for molecular clock dating and phylogenetic inference using ancient mitogenomes

  Li, W.; Sharif, B.; Heintzman, P. D.; Dalen, L.; Chacon-Duque, J. C.

  Abstract

  Summary: Ancient DNA studies are increasingly targeting samples that are beyond the limit of radiocarbon dating (>50 thousand years old) and are often difficult or impossible to date using other geochronological methods. In cases where complete mitochondrial genomes (mitogenomes) can be recovered from such samples, Bayesian molecular clock dating approaches are routinely used as an alternative method for estimating their age. However, molecular clock dating of ancient mitogenomes lacks a standardised, reproducible computational framework, and existing approaches rely heavily on graphical interfaces that limit automation and scalability. To address these gaps, we developed MitoDate, an automated Nextflow pipeline for reproducible molecular clock dating of ancient mitochondrial genomes. The workflow standardises Bayesian time-calibrated phylogenetic inference within a portable, containerised framework, reducing manual intervention and improving analytical consistency. Availability and implementation: MitoDate is implemented in Nextflow and is freely available at https://github.com/CpgSthlm/MitoDate. The pipeline is distributed with containerised dependencies and detailed documentation, including example datasets and usage guidelines.

  bioinformatics2026-07-31v1

• CyFj11: FlowJo v11 Workspace Import and Legacy Format Export for R-Based Flow Cytometry Analysis

  Jagla, B.; Culina, S.; Le-Guerroue, F.; Karkeni, E.; Hasan, M.

  Abstract

  High-dimensional flow cytometry measures immune cells at single-cell resolution, enabling systematic characterization of cell populations at scale. But harnessing this potential requires seamless interoperability between the interactive gating tools used by biologists and the statistical environments used for detailed downstream analysis. FlowJo, one of the most widely used commercial cytometry analysis software packages, now stores workspaces in a format that existing R tools cannot read, leaving researchers unable to import their gating strategies into R, or to return R-based results to FlowJo for visual review or collaborative sharing, without manual reconstruction. We present CyFj11, an R package that closes this gap, enabling import of FlowJo v11 gating hierarchies into R and export of R-defined gates back to FlowJo (throughout this paper, "import" refers to bringing a FlowJo v11 workspace into R, and "export" to writing an R-derived GatingSet back out to FlowJo). Using a combination of synthetic test scenarios and a real-world immunophenotyping dataset, we show that population counts in FlowJo 10 and 11 matched R-derived values with Pearson correlation coefficients exceeding 0.99. CyFj11 is platform-independent, requires no additional software infrastructure, and is freely available at https://github.com/C3BI-pasteur-fr/CyFj11.

  bioinformatics2026-07-31v1

• ChemoCalib: multiblock PLS calibration of genome-scale metabolic models improves flux prediction over expression-only integration

  zhang, X.

  Abstract

  Constraint-based metabolic modeling (FBA) needs multi-omics integration to narrow flux solution spaces, yet existing expression-only methods (E-Flux, GIMME, iMAT, MOMENT, E-Flux2, SPOT) ignore cross-omics covariance and rely on heuristic thresholds. We present ChemoCalib, a chemometrics-calibrated constraint layer that projects shared latent structure from metabolomics, transcriptomics, and proteomics onto genome-scale reaction bounds via multiblock partial least squares (MB-PLS) with GPR-aware VIP aggregation. Unlike linear expression scaling, ChemoCalib derives continuous, interpretable bounds (soft/hard/adaptive modes) and couples them to an in-silico active-learning loop for virtual double-knockout selection. On E. coli iJO1366, using three public multi-omics datasets (Ishii 2007, Keio, Holm 2010) and 13C-MFA ground truth, ChemoCalib achieves Spearman {rho} = 0.461 and Pearson r = 0.493 against measured fluxes, significantly outperforming E-Flux, GIMME, iMAT, MOMENT, E-Flux2 and SPOT (Holm-corrected p < 0.05 on 3/5 baselines). On the full 20-condition E-Flux2/SPOT curated set (~430 flux measurements, Supplementary Table S3), the gain is retained ({rho} = 0.48, p < 0.01 vs SPOT). FVA solution space contracts 55% with 94.2% wild-type feasibility preserved. A Gaussian-process surrogate quantifies uncertainty (reliability diagram ECE = 0.032). Pentose phosphate path routing shift ({Delta}{rho} = +0.088) aligns with known G6PDH post-transcriptional regulation. ChemoCalib is open source (Zenodo DOI: 10.5281/zenodo.21645890), needs no wet experiment, and positions chemometrics as a calibration layer for virtual-cell metabolic modules.

  bioinformatics2026-07-31v1

• AI4Life Open Calls and Public Challenges: why, how, and what we have learned.

  Galinova, V.; Seifi, M.; Serrano Solano, B.; Lidayova, K.; Dalle Nogare, D.; Corbat, A. A.; Talks, J.; Giacomello, E.; Gomez-de-Mariscal, E.; Ferreira, M. G.; Fuster-Barcelo, C.; Battagliotti, J. M.; Garcia-Lopez-de-Haro, C.; Salmon, B.; Croft, M.; Yie, S. Y.; Rey-Paniagua, G.; Hu, X.; Cho, S.; Sheth, A.; Porwal, C.; Li, X.; AI4Life Consortium, ; Henriques, R.; Li, X.; Krull, A.; Klemm, A.; Munoz Barrutia, A.; Kreshuk, A.; Ouyang, W.; Jug, F.; Deschamps, J.

  Abstract

  Within AI4Life, we ran three Open Calls and three Public Challenges (2023-2025), supporting 22 bioimage analysis projects from 151 applications and engaging 225 challenge participants, with the aim of applying FAIR deep learning in the life sciences. Our experience offers a view of the current state of bioimage analysis, the landscape of available tools, as well as the existing gaps between method developers, tool producers and potential users. It highlights that even after careful selection for AI-ready projects, most still require substantial effort to apply deep learning, and that the field still relies heavily on established, well-rounded methods to solve common problems. We come to the conclusion that for scientific AI in biology, the rate-limiting step is not methods and models but data, annotations, and shared infrastructure underneath them.

  bioinformatics2026-07-30v2

• General-purpose language models integrate structured biological evidence for explainable biological interaction prediction

  Zhang, Y.-z.; Xu, L.; Imoto, S.

  Abstract

  Biological interaction inference often requires integrating heterogeneous evidence that differs in biological meaning, provenance, specificity, and reliability. Existing computational approaches typically compress such evidence into numerical features, aggregate scores or latent representations before prediction, obscuring the contribution of individual evidence records and limiting interpretation when evidence is conflicting or incomplete. Phage--host interaction prediction provides a demanding test case for this problem, requiring the integration of genomic, functional, and reference-derived evidence with differing coverage, specificity, and reliability. Here we introduce a structured-evidence inference paradigm that retains heterogeneous biological observations as modular, named, and experimentally perturbable records rather than compressing them before prediction. We implement this paradigm in PHI-Reason, which constructs structured evidence profiles from genomic annotations, receptor-binding protein homology, nucleotide-neighbour relationships, alignment-free genomic similarity, and available CRISPR spacer information. A general-purpose large language model (LLM) performs inference directly over these structured profiles without task-specific training or manually engineered evidence-fusion rules. To validate this paradigm, we evaluated it across two distinct biological interaction domains: prokaryotic phage--host prediction and eukaryotic virus--host prediction. Across phage--host benchmarks, PHI-Reason achieved species-level top-1 accuracies of 63.6% and 53.2% on RefSeq-634 and VHDB-3150, respectively, and a multi-host accuracy of 0.571 on the Hi-C cohort, outperforming established numerical methods. Besides, systematic evidence perturbations quantified how individual evidence supported, complemented, or misled inference, identifying nucleotide-neighbour context as the dominant signal. Analyses of LLM intermediate representations using a target-conditioned local Jacobian readout, together with output-grounding analyses, further characterized evidence-dependent inference and quantified the extent to which generated rationales departed from the supplied evidence profiles. Applying the same framework to eukaryotic virus--host prediction using domain-appropriate evidence also preserved high predictive accuracy (66.6\%), supporting the generalizability of the proposed paradigm across biological prediction. These results demonstrate that LLMs can serve as an evidence-grounded inference interface, integrating heterogeneous biological evidence while making the boundaries of evidential support directly testable.

  bioinformatics2026-07-30v2

• An integrated single-cell atlas of human lung across the lifespan

  Huang, L.; Huang, Z.; Feng, S.; Fang, K.; Wu, J.; Ao, Y.; Huang, L.; Zhang, J.; Sun, H.; Miao, Z.

  Abstract

  The healthy adult lung is largely quiescent but can respond to injury and replace damaged or lost cells. However, the identities and origins of regenerative cell states remain poorly understood. Here we present the Human Developmental Lung Cell Atlas (HDLCA), a curated single-cell reference atlas of the respiratory system across the lifespan, integrating 253 datasets from 225 studies, and comprising over 18 million cells from 3,198 samples and 2,460 individuals across 34 anatomical locations in both health and disease. The HDLCA defines 142 consensus lung cell types, including 13 rare and 8 previously undescribed ones. Notably, we identify a rare intermediate alveolar epithelial progenitor (Int AP) population in normal lungs, with transcriptional signatures shared by alveolar type 2 (AT2) and alveolar type 1 (AT1) cells. Trajectory inference indicates that Int AP cells arise from both AT2 and SCGB1A1+ secretory cells and give rise to AT1 cells, with Hippo-YAP/TAZ signalling governing lineage specification. Int AP cells are associated with genetic susceptibility to idiopathic pulmonary fibrosis and chronic obstructive pulmonary disease, and their dysfunction may contribute to abnormal alveolar repair and regenerative failure in both diseases. The HDLCA provides a unified atlas-level reference for mapping respiratory cellular diversity, uncovering reparative cell dynamics and informing regenerative and cell-based therapeutic strategies.

  bioinformatics2026-07-30v2

• Determining gene specificity from multivariate single-cell RNA sequencing data

  Swarna, N. P.; Booeshaghi, A. S.; Rebboah, E.; Gordon, M. G.; Kathail, P.; Li, T.; Alvarez, M.; Ye, C. J.; Wold, B. J.; Mortazavi, A.; Pachter, L.

  Abstract

  An important application of single-cell genomics experiments is to identify genes specific to biological categories or experimental conditions. Although numerous approaches have been proposed to identify such genes, we consider an axiomatic approach based on defining properties that a specificity measure should have. This leads us to develop ember (Entropy Metrics for Biological ExploRation), which we show is the only method satisfying four key desired properties for a specificity measure. Applying ember to eight tissues from eight founder mouse strains, we find that gene specificity is often unintuitive: canonical markers can be supplanted, housekeeping genes are context-dependent, and mouse strain can drive unexpected cell type switching. Unsupervised learning on entropy metrics uncovers shared genes specialized to male gonads and kidney, as well as genes specific to non-consecutive developmental stages in the kidney. To facilitate further exploration of gene specificity in mice, we have also developed a comprehensive specificity database, along with a web interface, API and MCP server. Extending ember to a human PBMC dataset collected from 255 diverse individuals, we find that variation in PBMCs is largely localized to classical monocytes. We also find genes with unique specificity by sex, age and ancestral background. Together, these applications establish ember as a powerful tool and provide a roadmap for elucidating the impact of human genetic variation using the murine model.

  bioinformatics2026-07-30v2

• Quantitative assessment of cell fate commitment in single-cell transcriptomics using scCS

  Kriukov, E.; Ivleva, E.; Baranov, P.

  Abstract

  Cell fate trajectory inference is one of the key downstream methods in single-cell RNA-sequencing data analysis. Multiple existing tools allow studying such and help identify continuous cell fate dynamics, important genes along the trajectory, and reconstruct the transcriptional states a cell passes to its estimated final state. These methods have been essential to study various biological systems, yet cell fate by itself currently presents mostly qualitative analysis, and existing tools do not allow to directly quantify the fate-related parameters, including commitment, transition speed, fate affinity and entropy. Such quantifications may be performed through multiple parameters and result in better understanding and description of cell fates. We present scCS (single-cell Commitment Scoring), a scverse-friendly Python framework for this problem. scCS introduces Discounted Future-Fate Propagation (DFFP), which models the source transition graph as a geometrically stopped random walk that can reach endpoint anchors or stop unresolved, with a user-defined finite expected graph horizon. For each cell, the resulting probabilities are separated into total fate reach, relative affinity among reached fates, entropy-based fate specificity and reach-supported resolved commitment, while Signed Ordering Flux independently quantifies local progression. scCS also provides endpoint-anchor, graph-coverage and horizon-sensitivity diagnostics, an instantaneous local-direction mode, standardized visualizations, gene-level analyses and replicate-aware comparisons across experimental conditions. Applications to pancreatic endocrinogenesis and neural crest-Schwann-cell differentiation illustrate the general framework. scCS converts an explicit biological fate hypothesis into auditable cell-, population- and replicate-level quantities without redefining the source dynamics.

  bioinformatics2026-07-30v1

• RKMR: A Rapid Kernel Machine Regression Framework for Optimal Marker Detection in Spatial Omics Data

  Seal, S.; Neelon, B.; Chakraborty, A.; Mattila, C.; Rubinstein, M.; Chung, D.; Angel, P.; Ghosh, D.

  Abstract

  High-throughput spatial omics technologies enable molecular profiling within intact tissue architecture, yet identifying concise, predictive, and biologically interpretable marker panels for cell types, tissue domains, and disease-associated tissue classes remains challenging. This limitation hinders the development of actionable panels for targeted validation and downstream translation. Existing pipelines rely largely on univariate differential-expression analyses, which ignore joint molecular structure and provide limited predictive insight. Multivariate machine-learning methods, including random forest, XGBoost, elastic net, and specialized single-cell panel-selection approaches, can capture predictive patterns but typically lack explicit spatial modeling and probabilistic feature selection, relying instead on model-specific importance scores or user-specified panel sizes. We develop rapid kernel machine regression (RKMR), a scalable framework for spatial-omics marker discovery that integrates nonlinear kernel modeling, spike-and-slab variable selection, and spatial dependence. RKMR uses automatic relevance determination (ARD) kernels and sparsity-inducing priors to capture nonlinear marker-outcome relationships and implicit feature interactions while producing approximate posterior inclusion probabilities (PIPs) that quantify model-based uncertainty in feature inclusion. To scale inference to large spatial datasets, RKMR combines low-rank kernel approximations with stochastic variational optimization. In simulations, RKMR consistently achieves higher AUPRC than competing methods across a range of molecular-signal and spatial-effect settings. Across spatial transcriptomics and scRNA-seq datasets, RKMR identifies parsimonious marker sets that recover reported cell-type signatures and reproducible tissue-layer markers. These results establish RKMR as a scalable and uncertainty-aware framework for translating high-dimensional spatial omics data into robust, experimentally actionable marker panels.

  bioinformatics2026-07-30v1

• Expanded unbiased population-genomic summary statistics in pixy

  Samuk, K.; Stone, M.; McAuley, E.; Bailey, N. P.; Lucas, G.; Plaza, M.

  Abstract

  Estimates of population-genetic summary statistics are often computed from variant-only VCFs, which commonly omit invariant sites (i.e. sites with only homozygous reference genotypes across all samples). However, these sites are critical for correctly estimating per site statistics such as nucleotide diversity ({pi}) and between population divergence (dxy). Our software pixy addressed this issue by providing support for computing statistics directly from "all-sites" VCFs that encode invariant positions explicitly (Korunes and Samuk 2021). pixy has since been widely adopted and used in a large variety of population genetic studies. Here we present a major update to pixy, which expands the original tool in four major areas. First, we provide implementations of new estimators of Watterson's {theta} and Tajima's D that are unbiased with respect to missing data. Secondly, we provide support for arbitrary ploidy and multiallelic sites. Third, we introduce a variety of optimizations, including multicore execution and a roughly order-of-magnitude reduction in per-worker memory footprint. Finally, we have broadly modernized our code base, test suite, and community contribution pathway. We validate these new features on simulated polyploid and multiallelic datasets, as well as two empirical datasets (one diploid and one autotetraploid), and benchmark the new multicore scaling. We position pixy relative to contemporary tools and discuss shared limitations of VCF-based estimators. By integrating support for diverse ploidies, allelic complexity, and genome-wide scale in one workflow, pixy provides the population genetics community with a straightforward tool for estimates of key summary statistics that are unbiased with respect to missing data. pixy remains completely open-source (MIT-licensed) and easily installable with the conda package management system.

  bioinformatics2026-07-30v1

• Rich structure alphabets enable highest accuracy protein search

  Edgar, R. C.

  Abstract

  Protein structure databases have grown from thousands of experimentally determined structures to hundreds of millions of AI-predicted models, creating an urgent need for search methods that combine high accuracy with practical scalability. Here, I present the third generation of Reseek, a protein structure search algorithm achieving the highest overall accuracy (median rank 1) according to diverse metrics among tested methods including DALI, Foldseek and TM-align. Improved accuracy is obtained by parallel sequence alignment of many discrete alphabets capturing primary, secondary and tertiary features, giving a combined space of ~10^22 possible states. Separate statistical models are optimized for family, superfamily and fold discrimination, respectively, revealing distinct combinations of features that characterize each level. Hundreds of query structures can be searched a against a multimillion-structure database on a server computer in minutes, making large-scale structure search at state-of-the-art accuracy practical on commodity hardware.

  bioinformatics2026-07-30v1

• Resolving Immune Lineage and Cell-State Heterogeneity in Human PBMCs via Mass Spectrometry-Based Single-Cell Proteomics

  O'Connor, S. A.; Gletten, R. B.; Sharma, R.; Jensen, Z. N.; Lovell, B.; Garcia-Mansfield, K.; Ghoda, L. Y.; Zhang, B.; Frankhouser, D. E.; Rockne, R. C.; Trent, J. M.; Marcucci, G.; Pirrotte, P.

  Abstract

  Single-cell proteomics (SCP) currently lacks validated benchmarking standards, and cell annotation often relies on transcriptomic proxies. Unsupervised clustering offers a proxy-free alternative, but its success depends on biological signal outweighing technical variation. In homogeneous samples this is achievable, but in heterogeneous populations, where closely related cell types differ only subtly, technical variation can dominate the clustering and obscure the biology needed for annotation. To address this, we developed an integrated experimental and computational pipeline for protein-level cell annotation and applied it to human PBMCs as an immune-cell test case. We isolated T cells, B cells, monocytes, and NK cells by negative-selection sorting to build a high-fidelity reference. In parallel, unsorted PBMCs from the same donor were processed on a cellenONE and acquired using label-free DIA on an Orbitrap Astral Zoom. Using the labeled reference, we systematically benchmarked normalization, imputation, and clustering methods to assess their effect on cell-type separation. Unsupervised analysis resolved functional subpopulations within each lineage, and a probabilistic SCP classifier trained on these annotations identified the corresponding cell types and states in the unsorted PBMC fraction, validating the pipeline on unenriched, heterogeneous samples. Together, this work delivers an analytically benchmarked SCP workflow that resolves immune lineage and cell-state heterogeneity in human PBMCs and provides a classifier-ready, protein-level reference for immune-cell assignment.

  bioinformatics2026-07-30v1

• How Bias Shapes the Leaderboard: Scoring Function Performance Under Scrutiny

  Graber, D.; Kopko, J.; Stockinger, P.; Nakandalage, R.; Kuhn, B.; Mishra, S.

  Abstract

  Structure-based scoring functions leveraging machine learning have recently demonstrated superior performance over classical scoring functions, particularly on virtual screening benchmarks. However, due to the fundamental differences between their underlying model principles and architectures, it remains unclear to what extent performance stems from an understanding of molecular binding or from exploitation of systemic biases. Thus, disentangling the factors underlying benchmark performance is essential for determining whether a scoring function will generalize to novel chemical space and succeed in prospective drug discovery. To address this need, we present a case study investigating the nature and impact of systemic biases on benchmark comparisons between different scoring function paradigms. By systematically analyzing the evaluation workflows of prominent models, we reveal pocket bias, a form of spatial coordinate frame leakage arising from static binding pocket extraction, which artificially inflates benchmark performance. To progressively eliminate these sources of bias, we benchmarked two selected graph neural network scoring functions against two minimalist machine learning models and a classical scoring function under four increasingly stringent evaluation levels, successively removing pocket bias, reducing structural data leakage, and finally evaluating on out-of-distribution (OOD) protein targets. Upon removal of pocket bias and structural data leakage, the performance of all machine learning models dropped substantially. When evaluated on out-of-distribution protein families, the classical baseline AutoDock Vina outperformed the machine learning models in five of seven virtual screening tasks and dominated the docking power evaluation. Our findings indicate that benchmark performance can be heavily shaped by evaluation design and dataset artifacts, potentially overshadowing algorithmic improvements. While the tested machine learning models remain heavily dependent on encountering familiar data distributions to achieve competitive results, AutoDock Vina demonstrated superior generalization capacity on OOD targets. This work underscores the critical need for rigorous, artifact-free benchmarking protocols to guide the development of truly prospective machine learning models for virtual screening.

  bioinformatics2026-07-30v1

• Ontology-guided harmonization enables unified discovery of public metabolomics studies within and across repositories

  Banerjee, S.; Jalan, P.; Chinhara, R.; Kalle, C.; Wangikar, P.; Jadhav, K.

  Abstract

  Public metabolomics repositories contain thousands of studies, but differences in metadata structure, vocabulary, and repository-specific terms still limit reliable search, comparison, and reuse within and across databases. Here we present HARMONY, an ontology-based framework and web platform that harmonizes study-level metadata and metabolite information across Metabolomics Workbench and MetaboLights studies. HARMONY resolves eight biological and analytical metadata nodes, including species, sample source, disease, analytical technique, separation method, ion polarity, ionization source, and mass analyzer type, while preserving the original deposited terms as evidence. A ninth node, metabolite identity, maps metabolite entities to RefMet across both repositories. HARMONY uses a two-step workflow: Multi-source extraction retrieves records missed by single-field lookups, and ontology mapping then converts repository-specific labels into shared query terms, substantially closing the cross-repository retrieval gap relative to raw matching. Across the full corpus, HARMONY increased cross-repository retrievability from 75.5% to 89.6%, yielding thousands of study-node retrievals and reconnecting studies that raw-text search would have left unreachable within their own repositories. Approximately 91% of Metabolomics Workbench and 85% of MetaboLights studies had at least six of the eight nodes harmonized. The resulting platform, available at https://omicsinharmony.in, supports ontology-aware search, metadata filtering, within- and cross-repository study comparison, and metabolite-level querying, with retrieval backed by machine learning encoders that map study metadata into shared representations of biological and analytical context. HARMONY provides the metabolomics community with a shared, traceable search interface for study discovery and comparison within and across public repositories.

  bioinformatics2026-07-29v2

• The limits of Bayesian estimates of divergence times in measurably evolving populations

  Ivanov, S.; Fosse, S.; dos reis, M.; Duchene, S.

  Abstract

  Bayesian inference of divergence times for extant species using molecular data is an unconventional statistical problem: Divergence times and molecular rates are confounded, and only their product, the molecular branch length, is statistically identifiable. This means we must use priors on times and rates to break the identifiability problem. As a consequence, there is a lower bound in the uncertainty that can be attained under infinite data for estimates of evolutionary timescales using the molecular clock. With infinite data (i.e., an infinite number of sites and loci in the alignment) uncertainty in ages of nodes in phylogenies increases proportionally with their mean age, such that older nodes have higher uncertainty than younger nodes. On the other hand, if extinct taxa are present in the phylogeny, and if their sampling times are known (i.e., `heterochronous' data), then times and rates are identifiable and uncertainties of inferred times and rates go to zero with infinite data. However, in real heterochronous datasets (such as viruses and bacteria), alignments tend to be small and how much uncertainty is present and how it can be reduced as a function of data size are questions that have not been explored. This is clearly important for our understanding of the tempo and mode of microbial evolution using the molecular clock. Here we conducted extensive simulation experiments and analyses of empirical data to develop the infinite-sites theory for heterochronous data. Contrary to expectations, we find that uncertainty in ages of internal nodes scales positively with the distance to their closest tip with known age (i.e., calibration age), not their absolute age. Our results also demonstrate that estimation uncertainty decreases with calibration age more slowly in datasets with more, rather than fewer site patterns, although overall uncertainty is lower in the former. Our statistical framework establishes the minimum uncertainty that can be attained with perfect calibrations and sequence data that are effectively infinitely informative. Finally, we discuss the implications for viral sequence datasets. In a vast majority of cases viral data from outbreaks is not sufficiently informative to display infinite-sites behaviour and thus all estimates of evolutionary timescales will be associated with a degree of uncertainty that will depend on the size of the dataset, its information content, and the complexity of the model. We anticipate that our framework is useful to determine such theoretical limits in empirical analyses of microbial outbreaks.

  bioinformatics2026-07-29v2

• WITHDRAWN: Metabolic Rewiring in Triple-Negative Breast Cancer: Systems Analysis of TCGA-BRCA Transcriptome Reveals Prognostic Hub Genes

  Chandrasekar, S.

  Abstract

  The author has withdrawn this manuscript because of methodological issues. Therefore, the author does not wish this work to be cited as reference for the project. If you have any questions, please contact the corresponding author.

  bioinformatics2026-07-29v2

• Accurate ab initio gene prediction in eukaryotes with Tiberius in multiple clades

  Gabriel, L.; Bruna, T.; Kaur, A.; Krishnan, A.; Ortmann, F.; Salamov, A.; Talbot, S.; Becker, F.; Krieg, R.; Wheat, C. W.; Grigoriev, I. V.; Stanke, M.; Hoff, K. J.

  Abstract

  Background: Eukaryotic genome annotation is currently bottlenecked by limitations in the generality, scalability or accuracy of computational methods and fewer than 20% of the genomes available at NCBI Datasets have associated gene annotations. Evidence-based pipelines such as BRAKER3 achieve high accuracy but require substantial extrinsic evidence and compute. Deep learning approaches have recently achieved large improvements in ab initio gene prediction accuracy. The end-to-end deep learning gene predictor Tiberius approaches the accuracy of evidence-based annotation on mammalian genomes without using any extrinsic evidence, but its published models were trained on mammals only, limiting its applicability across eukaryotes. Results: We extend Tiberius beyond mammals by training lineage-specific models for Mesangiospermae, Fungi, Vertebrata, Insecta, Chlorophyta and Bacillariophyta, making the tool applicable to 92% of currently available eukaryotic assemblies. Across a benchmark of 33 species, Tiberius achieved higher exon-, transcript- and gene-level accuracy than the other evaluated ab initio methods, Helixer and ANNEVO, improving gene-level F1 score by 12-37 percentage points over Helixer and by 10-22 percentage points over ANNEVO, while also having the fastest runtimes overall. Compared with BRAKER3, which incorporates RNA-Seq and protein evidence, Tiberius approaches state-of-the-art accuracy in Mesangiospermae, Fungi, Bacillariophyta and Chlorophyta, while being on average 80 times faster when using a GPU. A reimplementation of the Tiberius backend reduced runtime by 31% compared to the previous implementation. Tiberius and its models are also available through a web server, which allows annotation of submitted assemblies without local resources. Furthermore, the Vertebrata model has been applied to annotate 2,948 vertebrate assemblies totalling nearly 6 trillion base pairs. Conclusions: Tiberius is transferable well beyond Mammalia and reaches accuracy close to evidence-based annotation in several clades at a small fraction of the compute cost. This makes it a practical choice for highly accurate large-scale genome annotation, particularly when extrinsic evidence is unavailable or annotation throughput is limiting. Availability and implementation: Code: https://github.com/Gaius-Augustus/Tiberius, web server: https://bioinf.uni-greifswald.de/tiberius.

  bioinformatics2026-07-29v2

• seqproc: An efficient, flexible, and concise tool for sequence geometry description and transformation

  Cape, N.; Fisher, E.; Liu, D.; Patro, R.

  Abstract

  Motivation: Novel sequencing technologies (for example, different single-cell chemistries and protocols) produce complex data. Often, the sequenced reads themselves encode critical technical information, such as the cell or molecule of origin. For effective pre-processing of this data and subsequent downstream analysis, it is required to efficiently and accurately identify, extract, and potentially normalize this information. Results: We introduce seqproc, a general-purpose sequence pre-processing tool based on a concise descriptive grammar to specify sequence matching and transformations. seqproc compiles a user-provided sequence geometry and transformation description into an execution graph, executed by the ANTISEQUENCE library. We demonstrate that \seqproc is faster on most chemistries, substantially more memory efficient, and at least as accurate as alternative tools that provide similar functionality, while having a more concise description syntax. Availability: seqproc is written in Rust and can be executed as a binary program or used as a Rust crate. It is licensed under the BSD 3-clause license and the source code is available at <a href="https://github.com/COMBINE-lab/seqproc">https://github.com/COMBINE-lab/seqproc</a>.

  bioinformatics2026-07-29v1

• Single-cell foundation modeling with species-nativeprotein tokens links regenerative competence across frog and mouse

  Cang, H.; Sun, S.

  Abstract

  Appendage regenerative capacity varies dramatically across developmental stages and anatomical sites, yet comparing these functional transitions across species remains difficult because cell types and gene vocabularies diverge. Universal cell embedding (UCE) provides a shared coordinate space across diverse organisms. Here we demonstrate that an experimentally anchored Xenopus laevis tail-organizer contrast identifies a developmental limb transition from which we define a frozen cross-species competence ruler. Applied without refitting, this ruler distinguished regenerative and fibrotic library groups in adult mouse digit repair and resolved a repeatable aligned component within state-balanced macrophage repair populations. Universal cell embeddings thus convert source population contrasts into portable vector rulers, enabling zero-shot functional comparisons across species boundaries.

  bioinformatics2026-07-29v1

• UDIST: unsupervised disentanglement of shape and texture for multi-scale phenotypic profiling in 2D microscopy

  Bosch, B. M.; Terpstra, M. L.; Smith, M. B.; van der Steen, K. H.; Jonker, C. T. H.; Ovcinnikovs, V.; Wesselink, T. H.; Janssen, A. F. J.; Winkel, L.; Huigen, E. M. A.; Lefferts, J. W.; Mastrobattista, E.; Elstak, E. D.; van den Berg, C. A. T.; Beekman, J. M.; van Beuningen, S. F. B.

  Abstract

  Microscopy-based phenotypic profiling relies increasingly on autonomous, unsupervised feature extraction, yet no existing method explicitly separates shape from texture into dedicated and independent latent subspaces by architectural design. Therefore texture, encoding critical biological information such as protein distribution and intracellular organisation, remains inaccessible as an independent feature domain in standard unsupervised approaches. This represents a fundamental limitation that prevents unbiased phenotypic analysis across biological scales. Here we introduce UDIST (Unsupervised Disentanglement of Shape and Texture), a sequential dual variational autoencoder (VAE) framework that tackles this fundamental limitation by explicitly decoupling shape from texture into independent, non-overlapping latent subspaces at the single-object level. By training two VICReg-regularised VAEs on principal-axis-aligned objects, UDIST separates binary shape from continuous texture information into rotation-invariant feature spaces, enabling separate downstream analysis of both domains. We validated UDIST across biological scales, from nuclei and single cells to patient-derived intestinal organoids, using both fluorescence and brightfield imaging, revealing phenotypic differences previously hidden by morphological variation and enabling the independent analysis of shape and texture in downstream analyses including clustering and similarity measurements. UDIST provides a versatile, label-free, and unsupervised tool for multi-scale phenotypic profiling in high-content microscopy and screening.

  bioinformatics2026-07-29v1

• Manifold Agentic Reasoning: Extending Agentic POMDPs and Post-Training Reasoning to Riemannian State and Reasoning Spaces

  Xu, T.; Hu, Z.; Sun, X.; Jin, L.; Xiong, M.

  Abstract

  Agentic reasoning systems increasingly interact with environments whose states are only partially observed, dynamically evolving, and constrained by physical, biological, or logical structure. Existing agentic reasoning frameworks often model internal reasoning, tool use, and post-training adaptation using flat latent representations and struggle in curved manifold space environments. However, many scientific and embodied domains naturally lie on curved state spaces, including tissue geometry, developmental trajectories, protein conformations, robotic configuration spaces, and constrained physical systems. We introduce Manifold Agentic Reasoning, a geometric framework that extends agentic reasoning from Euclidean latent spaces to Riemannian manifolds. In the proposed framework, observations are encoded as manifold-valued states, memory is retrieved by geodesic similarity, candidate hypotheses are generated in tangent spaces, predicted transitions are projected by exponential maps, and decisions are admitted through verification-gated commitment or repaired by manifold self-correction. We further extend the framework to graph-agentic manifold reasoning, where node states live on manifolds and neighbor information is transported by logarithmic maps before attention-based aggregation. Manifold agent reasoning moves AI past brittle, prompt-chained templates to solve four critical production flaws: silent hallucinations and reasoning drift, brittle tool and context misuse, the black-box evaluation problem and stiff behavior profiles. To evaluate the framework, we introduce a Curved Tissue Manipulation and Recovery benchmark in which an agent must repair damaged tissue on a curved manifold. Simulated results show that the full manifold-agent substantially outperforms both a baseline reasoning agent and a full flat-agent reasoning system, achieving higher recovery success, lower geodesic shape error, lower pattern error, and fewer invalid transitions. Curvature and ablation studies indicate that the performance gain is driven by geometry-aware reasoning, verification, memory, and self-repair. These results suggest that manifold-aware agentic reasoning provides a principled foundation for reliable scientific and embodied AI systems operating in curved, constrained, or mechanistic domains.

  bioinformatics2026-07-29v1

• Systematic cross-study assessment of RNA-Seq experimental workflows for plasma cell-free transcriptome profiling

  Tuni-Dominguez, C.; Asole, G.; Monteagudo-Mesas, P.; Rusu, E. C.; Cabus, L.; Gonzalez, L.; Sanchez, L.; Neto, B.; Sanders, P.; Weber, M.; Lagarde, J.

  Abstract

  Plasma cell-free RNA (cfRNA) is a promising source of non-invasive biomarkers, but its clinical translation is hindered by technical challenges and a lack of protocol standardization, which compromises reproducibility and comparability across studies. There is a need for a systematic evaluation of existing cfRNA-Seq workflows to understand the drivers of technical variability. Here, we address this gap by performing a comprehensive cross-study analysis of 2,166 cfRNA-Seq samples from 15 published studies and an in-house generated dataset, applying a uniform bioinformatics pipeline to enable a controlled comparison of experimental workflows. Our analysis reveals that the donor phenotype typically explains a negligible fraction of the transcriptomic variation, whose main determinants are technical -- principally protocol choice, genomic DNA contamination levels and library diversity. Remarkably, this technical noise is so profound that variation within plasma cfRNA samples exceeds that found across a wide range of human tissues. Finally, we demonstrate that critical pre-analytical factors are often confounded with patient phenotypes, jeopardizing the validity of biomarker discovery efforts. Our work serves as a comprehensive benchmark of current cfRNA-Seq methodologies and provides evidence-based guidelines to improve experimental design. By highlighting the dominance of controllable technical factors, we offer a path towards more robust and reproducible cfRNA research.

  bioinformatics2026-07-28v5

• Melody: Decoding the Sequence Determinants of Locus Specific DNA Methylation Across Human Tissues

  Jin, J.; Wang, D.; Qiao, J.; Gao, W.; Liu, Y.; Chen, S.; Zou, Q.; Wu, S.; Su, R.; Wei, L.

  Abstract

  DNA methylation is a fundamental epigenetic modification that plays crucial roles in transcriptional regulation, cellular differentiation, and genome stability. However, how locus-specific DNA methylation is determined by intrinsic DNA sequence remains poorly understood. Here, we introduce Melody, a deep learning framework that predicts DNA methylation from 10-kb genomic sequences, enabling the integration of both local and long-range sequence signals. Across 39 human tissues, Melody accurately predicts methylation profiles and consistently outperforms existing state-of-the-art methods in whole-chromosome, hypomethylated-region, and cell-type-specific benchmarks. Melody also generalizes to methylation quantitative trait locus (meQTL) effect prediction and identifies regulatory sequence motifs associated with methylation variability. To extend prediction beyond profiled tissues, we further develop Melody-G, which incorporates single-cell RNA-seq foundation model embeddings to infer methylation states in previously unseen cell types directly from transcriptomic data. Together, Melody provides a unified framework for linking genomic sequence and cellular state to DNA methylation and offers new insights into the regulatory logic governing the human methylome.

  bioinformatics2026-07-28v3

• Identifying spatially variable genes by projecting to morphologically relevant curves

  Nicol, P. B.; Ma, R.; Xu, R. J.; Moffitt, J. R.; Irizarry, R. A.

  Abstract

  Spatial transcriptomics enables high-resolution gene expression measurements while preserving the two-dimensional spatial organization of the biological sample. A common objective in spatial transcriptomics data analysis is to identify spatially variable genes within predefined cell types or regions within the tissue. However, these regions are often implicitly one-dimensional, making standard two-dimensional coordinate-based methods less effective as they overlook the underlying tissue organization. Here we introduce a methodology grounded in spectral graph theory to elucidate a one-dimensional curve that effectively approximates the spatial coordinates of the examined sample. This curve is then used to establish a new coordinate system that reflects tissue morphology. We then develop a generalized additive model (GAM) to estimate spatial patterns which permits the detection of genes with variable expression in the new morphologically relevant coordinate system. Our approach directly models gene counts, thereby eliminating the need for normalization or transformations to satisfy normality assumptions. A second important advantage over existing hypothesis-testing approaches is that our method not only improves performance but also accurately estimates gene expression patterns and precisely pinpoints spatial loci where deviations from constant expression occur. We validate our approach through extensive simulation and by analyzing experimental data from multiple platforms such as Slide-seq and MERFISH. As an example of its ability to enable biological discovery, we demonstrate how our methodology enables the identification of novel interferon-related subpopulations in the mouse mucosa, as well as markers of inflammation-associated fibroblasts in a multi-sample spatial transcriptomic dataset.

  bioinformatics2026-07-28v3

• ProNA3D: Distance-Based Analysis of Nucleic Acid-Containing Interfaces

  Genz, L. R.; Topf, M.

  Abstract

  Biomolecular interactions are central to many essential cellular processes, but RNA-containing complexes remain challenging to resolve structurally, even as experimental methods and AI-based prediction have expanded structural coverage. Tools for the integrated analysis of complex interfaces remain limited. We present ProNA3D, a tool that provides a unified platform for analyzing protein-nucleic acid and nucleic acid-only complexes, bridging the gap between structure prediction and functional interpretation. ProNA3D supports both experimental and computationally predicted structures, incorporating scoring metrics for AlphaFold3 predictions. It also offers interactive two-dimensional interface visualization and secondary-structure topology plots for RNA and DNA. An interface-based density zoning feature facilitates structure analysis in cryo-EM maps, allowing evaluation of dynamic complexes in the context of heterogeneous density. We demonstrate ProNA3D on diverse complexes solved by X-ray crystallography or cryo-EM, as well as on computational models. For example, in a trimeric complex of HIV-1 RNA and a human antibody, ProNA3D identified a high-connectivity nucleotide with potential functional relevance. Applying ProNA3D to the entire Protein Data Bank revealed distinct interface connectivity trends and interaction modes characteristic of specific complexes (e.g., in methyltransferase-DNA and CRISPR-associated) in nucleic acid-containing interfaces. The method is available as both a UCSF ChimeraX plug-in for visualization and a command-line tool at https://gitlab.com/topf-lab/ProNA3D. In addition, the repository contains the results of the large-scale analyses.

  bioinformatics2026-07-28v2

• ProtSyntax: a protein large language model for decoding post-translational modification syntax and function

  Lin, Y.; Wu, J.; You, Z.; Ni, X.; Chang, S.; Ding, J.; Wang, Y.; Gao, X.; Yang, S.

  Abstract

  Post-translational modifications (PTMs) expand protein function by encoding context-dependent regulatory states, and their dysregulation contributes to cancer, neurodegeneration and metabolic disease. However, existing methods treat PTMs as independent residue labels, limiting their ability to distinguish contextually permissible sites, model crosstalk and infer functional consequences. Here we introduce ProtSyntax, a PTM-aware foundation protein language model combining protein-aware positional encoding, bidirectional state-space propagation, geometry-constrained attention and adaptive multi-objective learning. This design integrates residue chemistry, motif order, long-range context and three-dimensional microenvironments while coupling PTM recognition to enzyme function. Across 40 PTM-site benchmarks, ProtSyntax exceeded the strongest baselines in mean MCC and AP by 12.66% and 10.67%. ProtSyntax also recovered masked PTM types and sites, rejected structural decoys, generalized to data-scarce modifications, reconstructed crosstalk and linked PTM perturbations to enzyme kinetics. Applications to pathogenic variants, biomolecular condensates and disease-associated PTM landscapes demonstrate its potential to decode the regulatory language of the modified proteome.

  bioinformatics2026-07-28v2

• TAMIPAMI: Software and methods for PAM/TAM identification in CRISPR and OMEGA gene editing systems

  Orosco, C.; Jain, P. K.; Rivers, A. R.

  Abstract

  Protospacer adjacent motifs (PAMs) and target-adjacent motifs (TAMs) are essential for target recognition by CRISPR-Cas and TnpB nucleases. Here we present TAMIPAMI, an efficient experimental and computational framework for rapid PAM/TAM identification. TAMIPAMI requires only a single control library and Cas or TnpB-treated library, simplifying experimental design, reducing cost, and providing greater accessibility for users. The platform interprets sequencing data with interactive visualizations and introduces a novel algorithm that determines the minimal exact set of degenerate IUPAC sequences describing the observed PAM/TAM patterns. Using this approach, we accurately recovered canonical motifs for several nucleases, including SpCas9, LbCas12a, AsCas12a, BrCas12b, Cas12i1, and AmaTnpB. TAMIPAMI is available as both a web application and command-line tool, ultimately providing an accessible and efficient platform for PAM/TAM discovery and characterization across CRISPR and OMEGA systems.

  bioinformatics2026-07-28v2

• Ordered Gromov-Hausdorff Metric: A New Tool for Comparative Analysis of Protein Structures

  Timofeev, A.; Anufriev, A.

  Abstract

  Motivation: Classical protein structure comparison metrics such as RMSD and TM-score assess geometric similarity via rigid-body superposition and sequence-dependent alignment, but they do not explicitly incorporate the linear order of amino-acid residues into the metric itself. The Gromov-Hausdorff (GH) metric compares metric spaces by their intrinsic shape, yet it also ignores order. Consequently, proteins with rearranged domains can appear geometrically similar under GH even when their evolutionary relationship is ambiguous. We introduce the Ordered Gromov-Hausdorff (OGH) metric, defined on linearly ordered metric spaces via optimal order-preserving isometric embeddings into a common host space, to rigorously incorporate residue order into structural comparison. Results: The theoretical OGH distance satisfies all metric axioms for finite ordered spaces (non-negativity, symmetry, identity of indiscernibles, and the triangle inequality), proved by standard metric-gluing arguments. For computation, we introduce an efficient surrogate based on a greedy monotonic alignment with an exponential order-penalty; we prove that any monotone alignment produced by the surrogate induces a monotone correspondence whose distortion rigorously bounds the theoretical distance from above (Lemma 2.2.5). The algorithm runs in O(n*w), where w is the search window width. Analytical properties include invariance under rigid-body motions, upper boundedness, Lipschitz continuity under small coordinate perturbations, and concavity in the weight parameter . On the Viral Affinity Dataset (28 viral proteins from HIV-1, SARS-CoV-2 and MERS-CoV), the practical OGH score increases monotonically with residue shuffling (up to 0.363 at 100% shuffling) and correlates strongly with TM-score (r=0.706). In the task of separating homologs at fixed global structural similarity (TM-score {approx}0.5), OGH achieves "AUC"=0.800, whereas TM-score gives "AUC"=0.467, demonstrating that order conservation is a more reliable signal of evolutionary relatedness than global geometry alone in the twilight zone of homology. Availability: The Python source code for OGH is freely available at https://github.com/andytimoffilim/OGH. The VAD dataset (PDB IDs listed in the paper) is publicly accessible from the RCSB Protein Data Bank [1,2].

  bioinformatics2026-07-28v2

• PanTEon: a cross-kingdom framework to guide the design of transposable element classifiers

  Orozco-Arias, S.; Ferrer-Pomer, I.; Rodrigues de Goes, F.; Gaviria-Orrego, S.; Gomiz-Fernandez, J.; Llatser-Torres, J.; Paschoal, A. R.; Guyot, r.; Gabaldon, T.

  Abstract

  Transposable elements (TEs) are major drivers of genome evolution, yet their annotation and classification remain inconsistent and hard to reproduce across species. Fragmented repeats, lineage-specific innovations, and heterogeneous taxonomies across databases and tools complicate comparisons and slow progress in TE biology. To address this, we developed PanTEon, a cross-kingdom deep learning framework for reproducible TE classification that combines a harmonized database with an open, modular benchmarking platform. The PanTEon Database is an automatically curated, taxonomically broad TE repository spanning animals, plants, and fungi. The PanTEon platform standardizes training, evaluation, and inference across nine Machine Learning methods, while remaining extensible to user-defined architectures. Using this framework, we benchmark state-of-the-art Machine Learning-based TE classifiers across TE superfamilies and major eukaryotic lineages and find that performance varies markedly by kingdom and superfamily. Ensemble approaches and phylum-specific models improve predictive F1 scores, but cross-species generalization remains a major challenge. Together, PanTEon Database and PanTEon platform provide a reproducible, scalable, and extensible foundation for TE classification, enabling standardized evaluation of future AI methods and supporting community-driven annotation efforts.

  bioinformatics2026-07-28v2

• SQUARNA: stem maximization for accurate de novo RNA secondary structure prediction

  Serdakov, M. D.; Bohdan, D. R.; Nikolaev, G. I.; Bujnicki, J. M.; Baulin, E. F.

  Abstract

  Non-coding RNAs play diverse roles in a wide range of cellular processes, with their spatial structure being pivotal to their function. RNA secondary structure is a key determinant of its overall fold. Given the scarcity of experimentally determined RNA 3D structures, understanding secondary structure is vital for discerning RNA function. Currently, there is no universally effective solution for de novo RNA secondary structure prediction. Existing methods are becoming increasingly complex without marked improvements in accuracy and often overlook critical features such as pseudoknots and alternative folds. Here, we introduce SQUARNA, a new approach to de novo RNA secondary structure prediction that is suitable for both individual RNA analysis and large-scale structural searches. SQUARNA revisits the concept of base pair maximization and develops it into a stem maximization idea coupled with the widely used free energy minimization (MFE) framework. SQUARNA can predict alternative structures and handle pseudoknots of arbitrary complexity. Benchmarking shows that SQUARNA outperforms existing methods, including deep learning models, in both single-sequence and alignment-based RNA secondary structure prediction. SQUARNA seamlessly integrates sequence and alignment information with experimental data, such as residue reactivities obtained by chemical probing, as well as other structural restraints, including automated searches for Rfam database templates, G-quadruplex patterns, and protein-binding motifs. SQUARNA is available as a standalone tool at https://github.com/febos/SQUARNA and as a web server at https://larnal.imol.institute.

  bioinformatics2026-07-28v2

• Pinc: a simple probabilistic AlphaFold interaction score

  Badonyi, M.; Toth-Petroczy, A.

  Abstract

  AlphaFold has enabled large-scale prediction of protein-protein and protein-nucleic acid complexes, but ranking and assessing the quality of predicted models remain challenging. Existing confidence scores are often highly parametrised and provide limited interpretability. We introduce a simple geometric framework that converts AlphaFold predicted aligned error (PAE) into conditional contact probability. We show that these probabilities are well calibrated to the fraction of native contacts observed across experimentally determined structures. Motivated by this, we define the Pinc score (Probability of interface native contacts) as the mean contact probability between interacting chains. Because the probabilistic interpretation extends to individual residues, Pinc captures local structural constraint beyond interfacial burial, enabling residue-level prioritisation of hotspot positions for mutational studies. Depending solely on a single empirically fixed contact radius, Pinc offers an interpretable path from PAE to interface confidence, matching or exceeding the classification performance of more complex methods across five independent benchmark sets. We provide a portable, dependency-free C program and a Google Colab notebook for calculating Pinc scores for AlphaFold models at https://git.mpi-cbg.de/tothpetroczylab/Pinc.

  bioinformatics2026-07-28v2

• A Preparation-Free Mixture-of-Experts Framework for Protein-Ligand Affinity Prediction

  Bao, H.; Dong, S.

  Abstract

  Protein-ligand affinity (PLA) prediction is central to AI-driven drug discovery, but precise interaction-based methods require costly conformation preparation and data encoding, limiting their throughput. To reconcile accuracy with efficiency, we first investigate whether pre-trained molecular representation models can replace complex encoders. A unified and diverse assessment of sequence-, graph-, and image-based representations reveals both strong overall performance and family-wise variability, delivering the first practical guidance for encoder selection in PLA tasks. Next, to achieve high computational efficiency without sacrificing expressiveness, we adopt the mixture-of-experts (MoE) strategy from large language models. Systematic ablation studies uncover key design principles for deploying MoE in molecular prediction. The resulting model, HydrAffinity, is an interaction-free, dynamic sparse model that uses pre-trained encoders and MoE for parameter-efficient learning. It outperforms all interaction-free methods and matches state-of-the-art interaction-based methods on CASF-2016. Routing analysis confirms that MoE develops distinct, family-specific activation patterns, providing interpretable evidence of dynamic parameterization across protein classes. Zero-shot tests on DUDE-Z and LIT-PCBA further show strong EF5% performance, making HydrAffinity a practical, scalable solution acting as an effective early-stage pre-filter.

  bioinformatics2026-07-28v1

• Image-based Disease-wide Association Study via Self-supervised Learning links Abdominal MRI Features to 158 Diseases

  Mousavi, S.; Arora, J.; Agarwal, R.; Fischer, A.; Mueller, S.; Ding, Z.; de Jong, J.; Yuan, K.

  Abstract

  Medical images contain rich phenotypic information that is often not fully captured by manual clinical assessment. Here, we present a systematic framework for extracting such information and conducting image-based disease-wide association studies (iDWAS) using self-supervised learning (SSL). We applied this framework to abdominal MRI data from the UK Biobank and evaluated associations between MRI-derived features and diseases. Features were learned using a VICReg-based SSL model and tested for disease associations using logistic regression models. We identified 158 diseases that were significantly associated with the MRI-derived features, including the ones which are not directly linked to abdominal anatomy. Focusing on metabolic dysfunction-associated steatohepatitis (MASH), we showed that the MRI-derived features captured disease-relevant information and separated MASH cases from controls better than established biomarkers such as PDFF and Iron-cT1. These findings highlight the ability of SSL to uncover clinically meaningful signals from routine imaging data. The proposed framework is broadly applicable to other imaging datasets and modalities, enabling more systematic approaches to incidental and early disease detection. The trained model and analysis pipeline are publicly available at https://github.com/srm2022/iDWAS.

  bioinformatics2026-07-28v1