New papers on Bioinformatics
174 new papers on bioinformatics in the last 7 days, within Biology. These are the 50 Pipette rates most worth reading, with the main result in the authors' own words.
The best of the week
Biomolecular design through latent reasoning with RFD4-Proteina
Here we introduce RFD4-Proteina, which addresses these limitations through intrinsic and extrinsic reasoning within a shared chemical and conformational latent space.
PreprintBold claims, read criticallyClaims a big stepReal-world useIs the Bitter Lesson coming to genomic prediction? An untuned tabular foundation model matches or beats specialized methods
Against the best of these, TabPFN-3.5 was significantly behind on only one trait, spruce wood density, and ahead on 12 of 18, with significant gains on all three soybean traits and rice flowering time.
PreprintBold claims, read criticallyA Biologically Interpretable Single-Cell Pathology Foundation Model for Molecular Decoding of H&E Images
Together, these findings establish Loki2 as a biologically interpretable single-cell pathology foundation model and provide the opportunities of recovering cellular and molecular information from routine H&E histology at single-cell resolution.
PreprintReal-world useQuantum Machine Learning for Missense Variant Pathogenicity Prediction in the KCNQ Gene Family
Here, we present, to our knowledge, the first application of quantum machine learning (QML) to missense pathogenicity prediction.
PreprintA hybrid biophysical-neural network model links nanoscale cardiac junctions to tissue conduction
Here, we show that a hybrid biophysical-neural network model removes this limit.
PreprintGenVinci reconstructs 3D molecular tissue scenes from histology to map senescence across aging and disease
GenVinci predicts spatial gene expression from H&E with zero-shot generalization and maps senescence-associated states across 85.8 million cells, revealing cell-type-specific patterns and disease-associated remodeling.
PreprintBold claims, read criticallyClaims a big stepDeepRNAi: Predicting on- and off-target siRNA activity and establishing a benchmark
We introduce DeepRNAi, a deep learning model that accurately predicts on- and off-target knockdown from siRNA and mRNA sequence.
PreprintBold claims, read criticallyReal-world useOptimal Predictivity of Random Signatures: Causes and Good Modeling Practices
The theory reveals as necessary and sufficient causes of these anomalous behaviors, the existence of large equivalence classes of optimally predictive feature sets, and their structure of information equivalences.
PreprintBold claims, read criticallyModBERT: a DNA modification and genetic variant-aware DNA large language model
Overall, ModBERT establishes a framework for personal-genome language modeling, in which a model reads each individual's genetic and epigenetic sequence rather than a shared reference.
PreprintCasanovo-TMT: deep learning-based de novo sequencing of TMTpro-labeled peptides and phosphopeptides
The resulting model, Casanovo-TMT, predicted the exact peptidoform for 84.7% of the TMTpro-labeled spectra in our test set and for 87.4 and 79.6% of the spectra in two public global-proteome datasets from other laboratories, whereas the extended vocabulary without fine-tuning gave 0.3%.
PreprintRevealing the platelet transcriptome using Siphon-seq: a sensitive, purity-aware method for RNA-sequencing
Combined, these results reveal that RNA-level contamination can alter both transcriptome structure and biological inference, and that Siphon-seq bridges the sensitivity-purity tradeoff that has limited platelet RNA sequencing, enabling sensitive, purity-aware studies of platelet biology in health and disease.
PreprintSegment Anything for Microscopy 2
Here we present Segment Anything for Microscopy 2 (SAM2), a foundation model that adapts SAM2 to joint 2D and 3D segmentation in light and electron microscopy.
PreprintBold claims, read criticallyAlphaGenome 2 predicts regulatory activity and variant effects across cell states
Here we present AlphaGenome 2 (AG2), which substantially advances cell-type-specific prediction of regulatory activity and variant effects for gene expression and chromatin accessibility.
PreprintGENECraft: A Function- and Prior-Constrained Discrete Dynamical System for Controllable Regulatory DNA Design
We introduce GENECraft, a training-free Function- and Prior-Constrained Discrete Dynamical System (FP-DDS) that couples a genomic language model (Carbon) and a functional oracle (AlphaGenome) during inference-time optimization to balance predicted regulatory activity and consistency with genomic sequence priors.
PreprintscUNVEIL: Learning scRNA-seq Cell Embeddings Through Measurement Enrichment
We introduce scUNVEIL, a self-supervised model trained through next-UMI prediction.
PreprintDeciphering the Biological Function of Tandem Mass Spectra Using a Dual-Modal Metabolite Language Model
Here, we introduce Spec2Function, a function-centric framework that infers biological functions from tandem mass spectrometry (MS2) spectra.
PreprintCell type-specific coexpression networks funnel genetic risk into core genes
Overall, accounting for shared housekeeping networks reveals that cell type-specific networks funnel genetic risk into downstream and dynamic core disease genes.
PreprintQuantifying Homology and Length Bias in Antimicrobial Peptide Prediction
Homology control costs 0.067 MCC and length matching a further 0.115, taking the reported score from 0.834 to 0.661.
PreprintChimera: Cross-Species Single-Cell Generative Modeling of Expression Deployment Relative to Protein Conservation
Chimera measures the second, a gene's deployment, on the same genes and reports the difference.
PreprintBold claims, read criticallyLearning gene co-regulation across cells with Multicellular Transformer
We present the Multicellular Transformer (MCT), a foundation model built around a different assumption: cells function together within tissues, so the appropriate unit of input is not one cell in isolation but a set of cells from the same sample.
PreprintBold claims, read criticallyReal-world usePallatom2: Navigating the De Novo Protein Universe with Atom-to-Atom Modeling
We introduce Pallatom2, an efficient, precise Atom-to-Atom generative model that jointly models protein sequence, backbone conformation, side-chain geometry, and target-interface contacts.
PreprintReal-world useAlgorithms for Designing Therapeutic Oligonucleotides on Pangenome Graphs
We provide a first methodological step towards principled, variant-aware design of therapeutic oligonucleotides.
PreprintReal-world useGene-scale DNA and chromatin ensembles with atomic resolution
Here, we introduce hierarchical chain growth for DNA (HCG-DNA) to build atomically detailed ensembles of double-stranded DNA and chromatin.
PreprintSavant: A Fast, Structure-Free Binding Affinity Predictor For Protein-Ligand Drug Screening
Leading computational methods co-fold the complex before scoring the binding affinity between the protein and the ligand, which provides high accuracy but requires a graphics processing unit (GPU) and substantial time for every pair evaluated.Here we present Savant, a machine learning model that estimates the dissociation constant using only the protein's amino acid sequence and the ligand's structure written as a line of text (SMILES).
PreprintBold claims, read criticallyReal-world useElflow: Automated Quality Control and Self-Supervised Representation Learning for High-Dimensional Cytometry
Here, we introduce Elflow, a modular, self-supervised framework that links automated quality control, single-cell representation learning, and patient-level modeling without manual gating at inference.
PreprintReal-world usepbzarr: cohort-scale storage and analysis of per-base genomic data with Zarr
Here, we present pbzarr, a format and toolkit that stores per-base genomic data from many samples and assays as a single Zarr store.
PreprintBold claims, read criticallyReal-world useCode availableMeasuring the information budget of cross-target antibody affinity generalization: a robust cross-target wall, and why structure reaches developability but not affinity
We introduce a noise-corrected, censoring-aware estimator of the cross-target transfer ceiling, {rho}max = {surd}(T/(T+I)), which separates a transferable antibody-intrinsic component (T) from a non-transferable target-specific interaction (I).
PreprintRevealing the transcriptomic organization of extrasomatic interfaces using SCOUBI
Here, we present SCOUBI, a computational framework that identifies axon-dendrite interfaces from extrasomatic RNA in ST data, revealing a hitherto overlooked layer of molecular function in neural tissues.
PreprintMicroDUET: a dual-headed genomic foundation model for pathogen surveillance
Here we present MicroDUET, which adapts a 26.6-million-parameter ProkBERT backbone with two 3.5-million-parameter adapter heads: a phylogeny head trained contrastively on genome windows from the same microbes, and a function head trained to identify genes across taxa.
PreprintBold claims, read criticallyReal-world useCharacterizing Multimodal Mutational Processes in Cancer Using MOTET
Applied to 2,583 whole genomes from the PCAWG cohort, MOTET recovers known MPs more accurately than standard post hoc correlation methods.
PreprintSMSTPred: Protein Language Model Embeddings Predict SAM-Methyltransferase Substrate Specificity and Reveal Metabolite Methyltransferasome Enrichment in Plants and Actinomycetota
Here we present SMSTPred (SAM-Methyltransferase Substrate-Type Predictor), a multi-label classifier that predicts MTase substrate type from protein language model embeddings and benchmarks it under homology-disjoint cross-validation against sequence-homology, structure-homology, and profile-HMM baselines.
PreprintscCody distinguishes aging-aligned from divergent disease-associated cellular remodeling
Here we introduce scCody, a generative framework that learns age-dependent cellular population dynamics from cross-sectional single-cell cohorts using population-level optimal transport and endpoint-consistency regularization. scCody accurately reconstructed cellular populations at unseen ages, recovered age-associated transcriptional changes more effectively than existing generative dynamics models, and enabled donor-age inference without direct age-regression training.
PreprintDynamic zonal identity is a functional determinant of liver disease
Our findings identify zonation as a dynamic component of disease phenotype and its restoration as a potential therapy.
PreprintBold claims, read criticallyReal-world useLearning spatial cellular state through 3D reconstruction of the cell
Here we present a representation learning framework that models each cell as a gene-annotated point cloud of transcripts and trains a rotation-equivariant neural network to reconstruct 3D cell shape, yielding spatially informed transcript embeddings.
PreprintIMPRESS: AI-coupled HPC workflows for protein design
We present IMPRESS (Integrated Machine-learning for PRotein Structures at Scale), a framework in which design pipelines are stateful objects that halt, continue, resample, or branch at runtime from their own accumulated results.
PreprintFlux Transformer learns intracellular metabolic fluxes from isotope labeling and measured reaction rates
Here we present Flux Transformer, an AI model that uses reaction-specific attention to integrate isotope-labeling data with available reaction-rate measurements, reconstructing mass-balanced steady-state fluxes.
PreprintBold claims, read criticallyAntibody design across observed and forecasted antigenic evolution
Together, these results establish multi-target computational design as a strategy for generating antibody breadth across observed antigenic variation and demonstrate the potential of evolutionary forecasting to guide antibody redesign against prospective escape.
PreprintReal-world useCIRCA learns identifiable cell- and niche-level programs and their interactions with tissue pathology from spatial transcriptomics
Here we present CIRCA (Causal Interaction Representation for Cellular Architectures), a self-supervised causal representation learning framework that jointly learns identifiable cell- and neighborhood-level transcriptional programs and the directed interactions among them under explicit assumptions.
PreprintSPIDER: Contextualizing Biomedical Knowledge for Evidence-Grounded Scientific Reasoning
Here we introduce SPIDER, a multi-agent framework grounding biomedical relations in literature evidence and extracting study conditions for interpretation.
PreprintspicyR enables robust, well-powered co-localisation tests in large spatial omics cohorts
We present spicyR v2, which avoids both problems with a fast, permutation-free test that accounts for these differences and weights each image by how precisely it measures co-localisation.
PreprintFoldMapper: Protein Inverse Folding Model with Structure-Derived Evolutionary Profiles and a Multiscale Graph Encoder
We present FoldMapper, an inverse folding model that breaks this circularity.
PreprintstatLens enables accurate and interpretable statistical intelligence
We present statLens, a model-agnostic AI-driven statistical intelligence framework that integrates LLMs, code and human review.
PreprintEvolutionarily Conserved Fragments reveal modular architecture and expand functional coverage in bacteriophage proteins
ECFs annotated 50% of phage representative proteins (at least doubling the coverage of either database), with the largest gains in structural and infection-related functions, a pattern replicated in independent metagenomic data.
PreprintDomain-informed effector gene assignment exposes training-data dependence of drug target prioritisation from GWAS
Performance did not transfer across domains (AUPRC 0.91 to 0.46 for the same method), and we traced this to how differently models weight the distance between variant and gene (p<5.3x10-123), not to how they were built.
PreprintCode availableInfernal-GPU: accurate E-values and sensitive covariance-model search for structured RNAs
We present Infernal-GPU, a GPU implementation of Infernal's CM dynamic programming that matches Infernal's results.
PreprintCode availableMultimodal protein representations across human brain cell types
To capture these complementary sources of information in the CNS, we introduce SEMPER, a multimodal representation learning model that aligns protein language model representations with protein-protein interaction networks from 442 cell types of the adult human brain, giving each protein a sequence-informed representation in every cell type in which its gene is expressed.
PreprintReal-world useBenchmarking Machine Learning Methods for Antimicrobial Peptide Prediction and Discovery in Ancient Microbiomes
Together, these findings establish a systematic benchmark for AMP prediction and highlight ancient microbiomes as an untapped source of antimicrobial peptide candidates.
PreprintFrom Screening to Differential Diagnosis: Benchmarking 3D Facial Representations Across Craniofacial Conditions
Overall, these findings support the potential of 3D facial representations for diagnostic candidate prioritization and retrieval, while emphasizing the influence of cohort composition on screening performance and the challenges of distinguishing phenotypically similar genetic syndromes.
PreprintReal-world useECMExplainer: a functional matrisome framework for decoding cellular stress states and aging trajectories
Here we introduce ECMExplainer, a framework that uses 180 matrisome gene-pair ratios, partitioned into 76 Assembly/Destruction and 104 Signaling/Signaling pairs to position fibroblast transcriptomes within a functional matrisome phenotype space defined by six reference fibroblast states spanning physiological and stress-associated conditions: Proliferating, Quiescence, Senescence, Fibrosis, Inflammation, and Hypoxia.
PreprintRNAtomy: A Structure-Aware RNA Foundation Model for Transferable Structural and Functional Representations
Here, we introduce RNAtomy, a self-supervised RNA foundation model designed to learn the structural anatomy of RNA from inter-nucleotide geometry and residue-level conformational descriptors.
Preprint