SEARCH · PubMed Health

Results for “probabilistic modelling”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 37 records · Page 2Linked to original sources

PubMed1977

[Taxonomy and assessment of psychological investigation methods in psychotropic drug trials (author's transl)].

Psychological tests are bound with specific goals. Four goal aspects are differentiated: status vs. processdiagnostics, normoriented vs. criterionoriented diagnostic, testing vs. inventarization, measurement of true scores vs. decision oriented diagnostic. Every diagnostic procedure is characterized by a specific personality theory and theory of measurement (classical vs. probabilistic test model). The diagnostic procedures traditionally used for evaluating drugs prefer status diagnostic, normoriented diagnostic, testing and measurement of true scores. Similar one-sidedness in personality theory and theory of measurement restrict validity and usefulness of psychological tests. In clinical practice we find a theoretically and empirically unjustified restriction in the selection of measurement devices on ratings and questionnaires. If we suppose multidimensionality of drug induced changes, we must apply a multimethod approach in outcome studies and use beside ratings and questionnaires behavior observations, objective tests, psychophysiological and neurophysiological measures. We propose a descriptive taxonomy of methods for planning multimethod outcome and process studies. From this taxonomy the methods of measurement for effects and side effects of drugs may be derived. The necessity of the multimethod approach is confirmed by empirical research. With this concept, the following neglected research questions become more obvious: concordance and discordance, synchrony and desynchrony of methods of measurement. We conclude with recommendations for clinical practice and research of outcome and process effects in drug therapy.

Clinical Trials as Topic↗

PubMed2026

Network methods for diagonal integration of unpaired single-cell multiomics data: a review.

MOTIVATION: Advances in single-cell sequencing have enabled multiomics profiling at unprecedented resolution; however, mass spectrometry-based single-cell proteomics (scMS) remains inherently destructive, precluding simultaneous transcriptomic capture. Unlike antibody-based methods such as CITE-seq, which permit paired profiling but are restricted to targeted protein panels, scMS provides unbiased, genome-scale coverage of the intracellular proteome yet necessitates post hoc integration of unpaired datasets. This diagonal integration challenge, where transcriptomes and proteomes are measured in separate cells lacking shared anchors, remains underserved by existing reviews, which focus predominantly on vertical integration strategies enabled by non-destructive assays. RESULTS: We survey the complete computational pipeline for constructing mechanistic proteogenomic networks from unpaired single-cell data, covering: (i) unimodal network inference such as knowledge-based approaches, probabilistic graphical models, temporal directionality inference, and generative and foundation model strategies that establish the transcriptomic scaffold; (ii) cross-modal integration architectures such as network propagation, graph neural networks (scMRDR, scmFormer, scCotag), and consensus frameworks designed explicitly for the unpaired proteomics setting; and (iii) benchmarking paradigms spanning network reconstruction (BEELINE, GRETA, CausalBench) and multi-task integration evaluation (scMultiBench, SCMMIB), with guidance on metric selection under network sparsity and class imbalance. We identify three principal axes of future development: generative proteomic translation from transcriptomic precursors, inductive prior embedding in next-generation architectures, and perturbation-based causal benchmarking. AVAILABILITY AND IMPLEMENTATION: This is a review article; no novel software is distributed. A curated benchmark resource table, methods starter guide, and per-method bottleneck annotations are provided in the Supplementary Material.

Multiomics↗

PubMed2026

A divide and conquer strategy for recapitulating whole genome 3D structure using Hi-C data.

The three dimensional (3D) spatial organization of the genome is closely linked to biological functions and can be captured by Hi-C assays through interrogating genome-wide chromatin interactions. Methodologies for inferring 3D structures from Hi-C data summarized as a two-dimensional (2D) contact matrix can be broadly placed within the paradigms of optimization-based and sampling-based. Many optimization-based methods are capable of constructing whole genome 3D structures but do not account for spatial dependency in the 2D data matrix nor cell heterogeneity in bulk Hi-C data, which provide an average over millions of cells. Sampling-based methods, on the other hand, are probabilistic model-based and can account for not only dependency, heterogeneity, but also other features inherent in Hi-C data, such as over-dispersion and sparsity. However, whole-genome 3D structure recapitulation is too computationally expensive for sampling-based methods, while chromosome-by-chromosome strategies for sampling-based methods ignore important information on inter-chromosomal contacts. To address these issues, we propose the truncated Random effect EXpression-cut and paste (tREX-cap) method, which applies the tREX model within a divide and conquer strategy. The resulting method inherits the good data-feature-cognizant properties of tREX and, in the meantime, can efficiently infer the whole genome 3D structure. We demonstrate the performance of tREX-cap through an extensive simulation study and analyses of a Hi-C lymphoblastoid dataset and a Hi-C IMR90 dataset.

Humans↗

PubMed2025

MutBERT: probabilistic genome representation improves genomics foundation models.

MOTIVATION: Understanding the genomic foundation of human diversity and disease requires models that effectively capture sequence variation, such as single nucleotide polymorphisms (SNPs). While recent genomic foundation models have scaled to larger datasets and multi-species inputs, they often fail to account for the sparsity and redundancy inherent in human population data, such as those in the 1000 Genomes Project. SNPs are rare in humans, and current masked language models (MLMs) trained directly on whole-genome sequences may struggle to efficiently learn these variations. Additionally, training on the entire dataset without prioritizing regions of genetic variation results in inefficiencies and negligible gains in performance. RESULTS: We present MutBERT, a probabilistic genome-based masked language model that efficiently utilizes SNP information from population-scale genomic data. By representing the entire genome as a probabilistic distribution over observed allele frequencies, MutBERT focuses on informative genomic variations while maintaining computational efficiency. We evaluated MutBERT against DNABERT-2, various versions of Nucleotide Transformer, and modified versions of MutBERT across multiple downstream prediction tasks. MutBERT consistently ranked as one of the top-performing models, demonstrating that this novel representation strategy enables better utilization of biobank-scale genomic data in building pretrained genomic foundation models. AVAILABILITY AND IMPLEMENTATION: https://github.com/ai4nucleome/mutBERT.

Humans↗

PubMed2026

Deep DNA and protein level feature integration for robust clinical variant interpretation using probabilistic gradient boosting.

A major challenge in clinical genomics is to classify genetic variations correctly, since it directly affects disease diagnosis and personal care. The existing methods tend to be based on the combination of different factors, such as protein structure, population frequencies, phenotypic annotations, and sequence conservation. Nevertheless, these methods often cannot be used to achieve the necessary interpretability, quantify uncertainty, and address rare cases. This paper presents a probabilistic gradient boosting model on variant pathogenicity prediction. The suggested framework applies biological characteristics at both level of DNA and protein levels while also scaling the level of uncertainty in clinical decision making. Our machine learning aims to solve the issues of variant interpretation by managing the features and through probability-based pathogenicity prediction. The framework formulation is aimed at generalizing over various datasets and minimizing overfitting. At the same time, it can ensure reasonable performance to facilitate clinical experiments. The model has also been tested on three standard datasets and demonstrated to be more predictive of the pathogenic effect of variants, in comparison with a variety of existing tools. The probabilistic gradient boosting model proposed had ROC AUC values of 0.9293, 0.9610, and 0.9646 on ClinVar variants, GRCh37, and GRCh38 human genome respectively. Furthermore, the dataset was ensured to include both exonic and intronic variants, and Variants of Uncertain Significance were also taken into consideration for Performance Testing. Through this it also aims to provide better clinical significance which will lead to a good interpretable tool for priority of variants for a large variety of disease conditions.

ClinVar↗

PubMed2026

ZILA-SRM: a probabilistic framework with zero-inflated latent models for robust strain reconstruction from metagenomes.

UNLABELLED: Resolving bacterial strain diversity from shotgun metagenomic data is fundamental to understanding intra-host evolution, transmission dynamics, and phenotypic heterogeneity. However, current probabilistic approaches face a severe "identifiability limit" when disentangling highly similar genomes. Under high-noise conditions, sequencing errors, coverage overdispersion, and collinearity confound standard expectation-maximization algorithms, resulting in overfitting and spurious "ghost" strains. Here, we introduce zero-inflated latent allocation for strain reconstruction from metagenomes with adaptive sparsity regularization (ZILA-SRM) to overcome this barrier through three innovations. First, we integrate a zero-inflated Poisson mixture model to decouple "structural zeros" (true strain absence) from "sampling zeros" (stochastic dropout), addressing overdispersion in standard Poisson-based tools. Second, we impose a convex adaptive sparsity regularization penalty that leverages biological sparsity priors to shrink noise artifacts dynamically. Third, we implement a graph-theoretic refinement step using maximal clique enumeration to resolve haplotype collinearity. Benchmarking against StrainFinder and MixtureS on 702 synthetic data sets shows that ZILA-SRM achieves a 20% improvement in precision in high-complexity scenarios while maintaining over 80% recall for minor variants at 0.5% abundance. Re-analysis of deep-sequencing data from 195 Mycobacterium tuberculosis clinical samples reveals cryptic low-abundance drug-resistant variants in 12% of patients, including a minor clone carrying the rpoB S450L mutation. Furthermore, application to skin microbiome data sets further reveals a strong negative correlation between dominant Staphylococcus aureus and Staphylococcus epidermidis strains, providing genomic evidence for competitive exclusion. These findings establish ZILA-SRM as a robust tool for resolving strain-level diversity in complex metagenomes. IMPORTANCE: Understanding microbial communities at the strain level is critical because closely related strains can differ dramatically in traits such as drug resistance, virulence, and ecological interactions. However, resolving individual strains from metagenomic sequencing data remains difficult, especially when strains are highly similar or present at low abundance. As a result, biologically meaningful diversity is often obscured or misinterpreted as noise. In this study, we introduce a new framework that improves the reliability of strain reconstruction from complex metagenomic data. By reducing false-positive strain detection while preserving sensitivity to rare variants, our approach enables more accurate characterization of microbial populations. This improved resolution reveals previously hidden subpopulations in clinical and microbiome datasets, providing clearer insights into microbial evolution, competition, and the emergence of clinically relevant traits such as antibiotic resistance.

Metagenomics↗

PubMed2026

BaGGLS: a Bayesian shrinkage framework for interpretable modeling of interactions in high-dimensional biological data.

MOTIVATION: Biological data is often high dimensional, noisy, and governed by complex interactions among sparse signals. This poses major challenges for interpretability and reliable feature selection. Tasks such as identifying motif interactions in genomics exemplify these difficulties, as only a small subset of biologically relevant features (e.g. motifs) are typically active, and their effects are often non-linear and context-dependent. While statistical approaches often result in more interpretable models, deep learning models have proven effective in modeling complex interactions and prediction accuracy, yet their black-box nature limits interpretability. RESULTS: We introduce BaGGLS, a flexible and interpretable probabilistic binary regression model designed for high-dimensional biological inference involving feature interactions. BaGGLS incorporates a Bayesian group global-local shrinkage prior, aligned with the group structure introduced by interaction terms. This prior encourages sparsity while retaining interpretability, helping to isolate meaningful signals and suppress noise. To enable scalable inference, we employ a partially factorized variational approximation that captures posterior skewness and supports efficient learning even in large feature spaces. In extensive simulations, we compare BaGGLS to frequentist probit regressions (unconstrained and with L1-penalty) as well as a probit model with Markov Chain Monte Carlo (MCMC) sampling under a horseshoe prior. We can show that BaGGLS outperforms the other methods with regard to interaction detection and is many times faster than MCMC sampling under the horseshoe prior. We also demonstrate the usefulness of BaGGLS in the context of interaction discovery from motif scanner outputs (e.g. Find Individual Motif Occurrences (FIMO)) and noisy attribution scores from deep learning models. This shows that BaGGLS is a promising approach for uncovering biologically relevant interaction patterns, with potential applicability across a range of high-dimensional tasks in computational biology. AVAILABILITY: Code is available at gitlab.com/dacs-hpi/baggls.

Bayes Theorem↗

PubMed2026

From genes to trajectories: mapping genetic influences on Huntington's disease progression.

MOTIVATION: There are many diseases with established genetic factors, such as Huntington's disease (HD), that are characterized by variable rates of progression. However, beyond the contribution of the known genetic factors - in this case the Huntingtin (HTT) gene - the impact of the full human genome on the natural progression of such diseases throughout a patient's life remains largely unknown. The increased availability of genome wide association (GWA) data in HD gene expansion carriers (HDGECs), combined with the clinical assessment scores on the same set of patients, has provided a perfect opportunity to assess the potentially broader genetic impact on the natural progression of HD. RESULTS: We present a genetics-driven, probabilistic disease progression model designed to identify and investigate the ways in which a range of genetic factors affect the natural progression of HD. When applied to a clinico-genomic HD dataset, our model identified several single nucleotide polymorphisms (SNPs) with previously unreported effects on disease progression that act at distinct stages and with varying magnitudes. This discovery may shed light on the potential mechanistic impact of previously unidentified genes on HD that may have implications for clinical management. As increasing amounts of GWA data become available more generally, we anticipate that this modeling framework will be broadly applicable to other diseases with strong genetic components. AVAILABILITY AND IMPLEMENTATION: The source code for IHDPM is available at https://github.com/BiomedSciAI/IHDPM.

Huntington Disease↗

PubMed2026

High-Purity Monovalent Functionalization of Carbon Nanotubes.

Single-walled carbon nanotubes (SWCNTs) show promise for probing molecular interactions at single-molecule resolution, yet generating SWCNT populations bearing a single defined functional tag remains challenging because surface functionalization is inherently stochastic. Here, we present a batch-scale strategy to produce predominantly singly tagged SWCNTs by leveraging the stochastic adsorption of single-stranded DNA (ssDNA). Specifically, SWCNTs are dispersed using a mixture of unmodified ssDNA (um-ssDNA) and a minor fraction of modified ssDNA (m-ssDNA) carrying an affinity handle. We developed a probabilistic ssDNA-SWCNT binding model that predicts the distribution of m-ssDNA per nanotube as a function of the input minor-strand fraction p = m-ssDNA/total ssDNA, enabling selection of conditions that maximize single-tag purity. Using magnetic-bead capture via a biotin affinity interaction and subsequent release, we isolate SWCNTs with 97.6% predicted single-tag purity at 2% recovery. Single-molecule fluorescence imaging further supports predominantly single-label occupancy under the model-selected conditions. Thus, this approach provides a general route to SWCNTs bearing a single molecular handle for downstream conjugation and assembly, supporting diverse future applications in SWCNT-based nanotechnologies.

Nanotubes, Carbon↗

PubMed2026

Accelerating inference in genomic and proteomic foundation models via speculative decoding.

MOTIVATION: Genomic and protein foundation models (GFMs and PFMs) have demonstrated strong performance in learning the language of DNA and proteins, but their use in large-scale sequence generation is limited by the latency of autoregressive decoding. Because every token triggers a forward pass of a large Transformer, whose inference is relatively slow, long-sequence generation quickly becomes costly. RESULTS: In this work we adapt speculative decoding to a representative GFM: the DNA model DNAGPT and two representative PFMs: ProGen2 and ProtGPT2. We implement a probabilistic variant of speculative decoding, in which a lightweight draft model proposes short token spans and a larger target model verifies or corrects them in parallel, while preserving the target model's sampling distribution. Across all three models we systematically study the effect of speculation window length, temperature, draft architecture and prompt length, and we benchmark tokens per second over multiple runs per configuration. Speculative decoding yields consistent speedups over standard key-value cached decoding, with maximum observed speedup reaching 100% increase, while average gains across models ranging between 20% and 40% (e.g. 1.2×-1.4×), without changing the underlying target model predictions. Our results show that speculative decoding is a practical and model-agnostic strategy for accelerating genomic and proteomic sequence generation without sacrificing prediction quality. AVAILABILITY AND IMPLEMENTATION: All code and results are freely available at https://github.com/Georgakopoulos-Soares-lab/BioSpecDec.

Genomics↗

PubMed2026

TPMM: three-component posterior mixture model enables robust inverton detection in low-depth metagenomes and suggests potential viral invertons.

SUMMARY: Bacterial phase variation enables reversible, locus-specific phenotypic switching, often driven by DNA inversion (invertons). To identify these events, researchers commonly rely on sequencing reads that provide orientation-specific support. Metagenomic sequencing, which captures total genetic material independent of cultivation, offers a powerful platform for the comprehensive study of invertons. However, computational inverton calling from metagenomic data is difficult at low sequencing depth: hard read-support cutoffs can miss true events, while sequence-only predictors lack read-backed interpretability and uncertainty quantification. To address this, we present TPMM, a three-component posterior mixture model for inverton calling in metagenomic data. TPMM explicitly incorporates sequencing depth to formulate inverton detection as a probabilistic mixture problem. Starting from candidates flanked by inverted repeats, the model classifies the candidates into noise, low-probability, or high-probability inversion signals using read evidence. Finally, TPMM assigns posterior probabilities as soft labels and applies cumulative Bayesian False Discovery Rate control to robustly identify true invertons. On two real gut metagenomic datasets, TPMM agrees well with PhaseFinder at high depth but recovers substantially more invertons under systematic downsampling, demonstrating superior performance in sparse-data regimes. We further examine potential reversible inversion elements in viral genomes and provide supporting analyses, suggesting a broader scope for inversion-mediated regulation. AVAILABILITY: The source code of TPMM is available via: https://github.com/KennyxxD/TPMM.

Metagenomics↗

PubMed2026

LAML-Pro: joint maximum likelihood inference of cell genotypes and cell lineage trees.

MOTIVATION: Recent dynamic lineage tracing technologies use genome editing to induce heritable mutations, or edits, that accumulate across successive cell divisions. These edits are measured using single-cell sequencing or imaging, providing data to reconstruct cell lineages at single-cell resolution. Current computational approaches to infer cell lineage trees, or phylogenies, from these data perform two separate steps: (i) Identify each cell's edits (genotype) from the raw sequencing or imaging data; (ii) Infer a cell lineage tree from the cell genotypes. However, genotyping cells is an inexact process and genotype errors can yield an inaccurate lineage tree. For example, using fluorescence based-imaging to measure edits results in a high fraction (≈25%-50%) of uncertain or erroneous genotypes. RESULTS: We introduce Lineage Analysis via Maximum Likelihood with PRobabilistic Observations (LAML-Pro), an algorithm that jointly infers cell genotypes and a cell lineage tree. LAML-Pro is based on the Probabilistic Mixed-type Missing Observation (PMMO) model, which we derive to describe both the genome editing and genotype observation processes. LAML-Pro constructs lineage trees from thousands of cells in under an hour by leveraging the sparsity of transitions under the PMMO model. On simulated data, we demonstrate that LAML-Pro corrects genotype errors and infers substantially more accurate trees than existing methods which are vulnerable to genotype errors. Applied to data from two recent imaging-based lineage tracing systems, LAML-Pro reduces genotype errors by 5-fold and produces more spatially coherent lineage trees compared to existing methods. AVAILABILITY AND IMPLEMENTATION: LAML-Pro is implemented in C++ and is available as both a command-line interface and as a Python library at: github.com/raphael-group/LAML-Pro.

Cell Lineage↗

PubMed2026

Deconvolution of evolutionary architecture unmasks a high-risk, subclonal-rich subtype in treatment-naive small cell lung cancer.

BACKGROUND: Intratumoral heterogeneity (ITH) drives therapeutic resistance in small cell lung cancer (SCLC). However, conventional single-sample analysis has limited horizontal, cross-patient comparisons, leaving the overarching evolutionary architecture in treatment-naive tumors poorly understood. This study aims to deconvolve these architectures to identify clinically relevant evolutionary subtypes. METHODS: We analyzed whole-exome sequencing data from 41 treatment-naive SCLC patients. To overcome the cross-patient comparability bottleneck, we developed a novel probabilistic framework using a refined Gaussian Mixture Model (GMM). This standardized subclonal structures into four hierarchical strata, enabling the identification of evolutionary subtypes via unsupervised clustering. To address the scarcity of SCLC public data, prognostic concordance was robustly explored in The Cancer Genome Atlas (TCGA) lung squamous cell carcinoma (LUSC) based on shared smoking etiology, with lung adenocarcinoma (LUAD) serving as a negative control. RESULTS: The cohort robustly segregated into "Clonal-dominant" (Group 1, n=28) and "Subclonal-rich" (Group 2, n=13) subtypes. Group 1 evolution was primarily driven by tobacco signatures (SBS4). Conversely, Group 2 exhibited late-stage acquisition of a DNA mismatch repair deficiency (MMRd) signature (SBS15), fueling trace subclonal diversification. Clinically, Group 2 demonstrated a significantly lower objective response rate (ORR) to platinum-based regimens (25.0% vs. 81.3%, P=0.02). Furthermore, the Subclonal-rich architecture independently predicted inferior overall survival (OS) [adjusted hazard ratio (adj. HR) =2.93, P=0.02], driven predominantly by limited-stage disease. Cross-cancer analysis validated this histology-dependent, high-heterogeneity adverse pattern in early-stage LUSC but not in LUAD. CONCLUSIONS: This hypothesis-generating study demonstrates that a "Subclonal-rich" architecture, driven by acquired MMRd, identifies high-risk, chemo-resistant SCLC. Our GMM approach suggests that pre-existing heterogeneity may serve as a potential, histology-dependent prognostic marker that warrants prospective validation for tailoring future therapeutic regimens.

Gaussian Mixture Model (GMM)↗

PubMed2026

LAML-Pro: Joint Maximum Likelihood Inference of Cell Genotypes and Cell Lineage Trees.

MOTIVATION: Recent dynamic lineage tracing technologies use genome editing to induce heritable mutations, or edits, that accumulate across successive cell divisions. These edits are measured using single-cell sequencing or imaging, providing data to reconstruct cell lineages at single-cell resolution. Current computational approaches to infer cell lineage trees, or phylogenies, from these data perform two separate steps: (1) Identify each cell's edits (genotype) from the raw sequencing or imaging data; (2) Infer a cell lineage tree from the cell genotypes. However, genotyping cells is an inexact process and genotype errors can yield an inaccurate lineage tree. For example, using fluorescence based-imaging to measure edits results in a high fraction (≈ 25-50%) of uncertain or erroneous genotypes. RESULTS: We introduce Lineage Analysis via Maximum Likelihood with PRobabilistic Observations (LAML-Pro), an algorithm that jointly infers cell genotypes and a cell lineage tree. LAML-Pro is based on the Probabilistic Mixed-type Missing Observation (PMMO) model, which we derive to describe both the genome editing and genotype observation processes. LAML-Pro constructs lineage trees from thousands of cells in under an hour by leveraging the sparsity of transitions under the PMMO model. On simulated data, we demonstrate that LAML-Pro corrects genotype errors and infers substantially more accurate trees than existing methods which are vulnerable to genotype errors. Applied to data from two recent imaging-based lineage tracing systems, LAML-Pro reduces genotype errors by 5-fold and produces more spatially coherent lineage trees compared to existing methods. AVAILABILITY AND IMPLEMENTATION: LAML-Pro is freely available at: github.com/raphael-group/LAML-Pro.

Journal Article↗

PubMed2026

Implicit and explicit statistical learning in reading: Evidence from a randomized controlled-learning study and computational modeling.

A key challenge in reading acquisition is understanding how learners extract the complex probabilistic mappings between print, meaning, and sound. Statistical learning (SL) theory offers a mechanistic account of how such mappings are acquired, whether implicitly through exposure or explicitly through instruction. We conducted a randomized controlled-learning study in Chinese, a writing system characterized by multiple sub-lexical regularities linking orthography, semantics, and phonology. Ninety-five 2nd-3rd graders with or at risk for dyslexia were randomly assigned to one of three groups: an implicit-SL training group exposed to repeated lexical and sublexical orthography-semantics-phonology associations, an explicit-SL training group receiving the same input plus explicit instruction on the sublexical print-sound mapping, and a no-SL control group. Both SL groups outperformed controls on the characters they were trained on, as well as on untrained characters that required generalization. However, only the explicit group demonstrated abstraction of print-sound mapping to novel items. Neural network simulations further revealed distinct mechanisms supporting implicit and explicit SL, consistent with a dual-system account of reading acquisition. Together, these findings (1) clarify how implicit and explicit learning distinctly support the discovery of statistical structure in written language and (2) underscore the implicit-explicit dual learning mechanism underlying reading acquisition.

Humans↗

PubMed2026

A token-pruning framework enables efficient representation of the human genome for RNA modification analysis.

MOTIVATION: Modelling long genomic sequences remains challenging due to extreme sequence length, high redundancy, and the need for biological interpretability. Although Transformer-based architectures have achieved strong performance across genomic tasks, their high computational cost and reliance on fixed tokenization strategies limit their scalability and ability to focus on biologically informative regions. RESULTS: We propose ATSFormer, a token-pruning Transformer framework for efficient and biologically informed genomic sequence modelling. ATSFormer incorporates an attention-guided and parameter-free Adaptive Token Sampling (ATS) module into Transformer layers. Guided by attention-derived importance scores, ATS dynamically retains informative tokens while probabilistically discarding redundant ones, thereby reducing sequence length, FLOPs, and memory usage without introducing additional learnable parameters or extra training procedures. Importantly, the retained tokens correspond to key contributors to model predictions, enabling ATSFormer to highlight biologically meaningful sites and sequence motifs. We evaluated ATSFormer on four benchmark RNA modification datasets derived from RMVar 2.0, covering A-to-I, m1A, m5C, and m7G. Experimental results show that ATSFormer consistently outperforms existing state-of-the-art methods while achieving substantial computational savings. Furthermore, structural analysis using AlphaFold3 supports the biological relevance of the motifs identified by ATSFormer. AVAILABILITY AND IMPLEMENTATION: The source data and code are freely available at GitHub (https://github.com/1gao2/ATSFormer) and Zenodo (https://doi.org/10.5281/zenodo.21813541).

Humans↗

PubMed2026

transfactor: transcription factor activity estimation via probabilistic gene expression deconvolution.

Gene expression is a primary modality being studied to differentiate between biological cells. Contemporary single-cell studies simultaneously measure genome-wide transcription levels for thousands of individual cells in a single experiment. While the characterization of cell population differences has often occurred through differential gene expression analysis, tiny effect sizes become statistically significant when thousands of cells are available for each population, compromising biological interpretation. Moreover, these large studies have spurred the development of methods to infer gene regulatory networks (GRNs) directly from the data, and GRN databases are becoming more comprehensive. In this work, we propose a statistical model for gene expression measures and an inference method that leverage GRNs to deconvolve transcription factor (TF) activity from gene expression, by probabilistically assigning mRNA molecules to TFs. This shifts the paradigm from investigating gene expression differences to regulatory differences at the level of TF activity, aiding interpretation and allowing prioritization of a limited number of TFs responsible for significant contributions to the observed gene expression differences. The inferred TF activities result in intuitive prioritization of TFs in terms of the (difference in) estimated number of molecules they produce, in contrast to other widely used methods relying on arbitrary enrichment scores. Our model allows the incorporation of prior information on the regulatory potential between each TF and target gene and is able to deal with both repressing and activating interactions. We compare our approach to other TF activity estimation methods using two simulation experiments and two case studies. Single-cell RNA-sequencing; TF activity; bioinformatics; GRN.

Transcription Factors↗

PubMed2026

Robust optimisation for photon radiotherapy: A scoping review of models, paradigms, and reporting.

BACKGROUND AND PURPOSE: Robust optimisation offers an alternative to conventional margin-based photon radiotherapy planning by explicitly modelling uncertainty, but practice is variable and not standardised. MATERIALS AND METHODS: A scoping review was conducted to map robust optimisation for photon external beam radiotherapy. Electronic searches of Scopus, PubMed and Google Scholar (2000-2025, English language) identified planning studies that incorporated modelled uncertainties into the optimisation process and reported at least one robustness-related outcome. Data were charted on clinical context, uncertainty models, optimisation paradigms, robustness metrics and evidence for clinical implementation. RESULTS: Seventy-one studies were included. Most investigated prostate, breast or lung cancer and used intensity-modulated radiotherapy or volumetric-modulated arc therapy in commercial or research treatment planning systems. Scenario-based worst-case (minimax) optimisation was the dominant paradigm in clinically oriented work, while chance-constrained, conditional value at-risk, distributionally robust and adaptive formulations were confined to small methodological series. Uncertainty modelling focused mainly on rigid set-up error; fewer studies incorporated respiratory motion, inter-fraction anatomical change, dose-calculation uncertainty or biological variation. Robustness was evaluated with diverse scenario-based dose-volume metrics, probabilistic coverage measures, composite robustness indices and, less often, biological endpoints. Direct clinical implementation reports were scarce. CONCLUSION: Robust photon planning is technically feasible and generally maintains or improves target coverage and organ sparing compared with margin-based planning. However, heterogeneity in uncertainty models, optimisation configuration and robustness reporting limits comparison and synthesis. Pragmatic minimum standards are proposed to support future consensus and wider clinical adoption.

Humans↗