SEARCH · PubMed Health

Results for “Dataset”

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 19 recordsLinked to original sources

PubMed2026

Dataset Readiness Assessment With Large Language Model (DRAFT-LLM): A Multi-Axis Audit Guided by LLM.

This article details the Dataset Readiness Assessment for Training (DRAFT), a systematic method for determining whether a high-dimensional biological dataset is suitable for developing reliable, equitable (i.e., the extent to which model performance, error patterns, and potential benefits or harms are evaluated and found to be acceptably distributed across relevant demographic, biological, clinical, and contextual subgroups), and scientifically meaningful machine-learning models, and DRAFT Large Language Model (DRAFT-LLM), its optional human-in-the-loop extension for calibrating study-specific audits through structured, critically reviewed LLM guidance. Standard model validation often fails to detect when apparent performance is driven by spurious correlations, technical artifacts, or hidden stratification, leading to irreproducible and inequitable findings. DRAFT-LLM addresses this gap by shifting the focus from model tuning to structured dataset auditing, organized around Support Protocols 1 to 4 that capture the scientific intent, data structure, and governance constraints of a given study. These Support Protocols: (1) elicit and formalize investigator input into a study intake and dataset card; (2) compute standardized dataset statistics and structural summaries suitable for downstream analysis and LLM context; (3) configure the language model using form-based responses, safety guardrails, and governance rules; and (4) generate personalized instructions, prompts, and code templates for running DRAFT audits. Basic Protocols 1 to 3 are instantiated from this support layer for generalization, equity, and stability: they are reusable execution patterns whose concrete behavior is determined by the cards, statistics, and configurations defined in the Support Protocols. DRAFT-LLM and DRAFT are demonstrated in this article through an end-to-end case study on The Cancer Genome Atlas (TCGA). © 2026 Wiley Periodicals LLC. Support Protocol 1: Study intake and dataset card construction Support Protocol 2: Dataset structure and advanced summary statistics for LLM context Support Protocol 3: LLM configuration using structured form responses Support Protocol 4: Generation of personalized instructions for DRAFT audits Basic Protocol 1: Generalization audit Basic Protocol 2: Equity audit Basic Protocol 3: Stability audit.

Large Language Models↗

PubMed2026

Freely available genomic datasets for atrial fibrillation research: current resources and analytical pipeline.

Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia, characterized by clinical and genetic heterogeneity. Increasing use of genomics and other omics approaches has driven reliance on publicly available AF datasets to advance biological discovery. Thus, this systematic review aimed to identify freely available genomic AF datasets through Mendeley Data and its interconnected repositories, and to characterize the most common analyses performed on these data. The search was conducted in adherence to the PRISMA 2020 guideline. Nineteen freely available genomic AF datasets were identified: Summary statistics for 'Biobank-driven genomic discovery yields new insight into atrial fibrillation biology', hum0014.v8.58qt.v1, AF GWAS in UK Biobank, UK Biobank (Publication 9659), GWAS summary statistics from a 2025 multi-ancestry AF meta-analysis, GSE115574, GSE128188, GSE14975, GSE2240, GSE238242, GSE254133, GSE261170, GSE271748, GSE271839, GSE293813, GSE294456, GSE31821, GSE41177, and GSE79768. The GEO datasets were further examined using differential gene expression, functional enrichment, protein-protein interaction networks, hub gene analysis, microRNA target prediction, and gene clustering, as well as, for the more recently deposited datasets, eQTL colocalization, single-cell/single-nucleus clustering, cell-cell communication analysis, and gene-dosage-dependent transcriptional and electrophysiological profiling. These analyses show some consistency but also considerable heterogeneity in initial conditions, data normalization, and analytical methodological settings. In conclusion, only a limited number of datasets are freely available, so additional, well-characterized and standardized datasets are needed to provide a complete picture of the AF pathology.

Mendeley Data↗

PubMed2025

Negative dataset selection impacts machine learning-based predictors for multiple bacterial species promoters.

MOTIVATION: Advances in bacterial promoter predictors based on machine learning have greatly improved identification metrics. However, existing models overlooked the impact of negative datasets, previously identified in GC-content discrepancies between positive and negative datasets in single-species models. This study aims to investigate whether multiple-species models for promoter classification are inherently biased due to the selection criteria of negative datasets. We further explore whether the generation of synthetic random sequences (SRS) that mimic GC-content distribution of promoters can partly reduce this bias. RESULTS: Multiple-species predictors exhibited GC-content bias when using CDS as a negative dataset, suggested by specificity and sensibility metrics in a species-specific manner, and investigated by dimensionality reduction. We demonstrated a reduction in this bias by using the SRS dataset, with less detection of background noise in real genomic data. In both scenarios DNABERT showed the best metrics. These findings suggest that GC-balanced datasets can enhance the generalizability of promoter predictors across Bacteria. AVAILABILITY AND IMPLEMENTATION: The source code of the experiments is freely available at https://github.com/maigonzalezh/MultispeciesPromoterClassifier.

Machine Learning↗

PubMed2025

Asymmetric integration of various cancer datasets for identifying risk-associated variants and genes.

MOTIVATION: Cancer genomic research provides an opportunity to identify cancer risk-associated genes, but often suffers from undesirable low statistical power due to a limited sample size. Integrated analysis with different cancers has the potential to enhance statistical power for identifying pan-cancer risk genes. However, substantial heterogeneity across various cancers makes this challenging. RESULTS: Recently, a novel asymmetric integration method was developed that can deal with data heterogeneity and exclude unhelpful datasets from the analysis. We adapted and applied this method to integrate genotype datasets with matched case and control individuals from the Michigan Genomics Initiative, using each cancer as the primary dataset of interest and the other cancers as auxiliary datasets, respectively. Conditional logistic regression models were coupled with the asymmetric integrated framework to handle the matched case-control study design and permutation tests were performed to control for false discovery rates (FDRs). At the same FDR level, the integrated analysis found more potential genetic variants and genes that are associated with the risks of various cancers, showcasing the promise of the proposed approach for integrated analysis of cancer datasets. AVAILABILITY AND IMPLEMENTATION: Our method is available as source code at https://github.com/rxxwang/integrate_cancer.

Journal Article↗

PubMed2026

usiGrabber: automating the curation of proteomics spectra data at scale, making large datasets ready for use in machine learning systems.

MOTIVATION: An unprecedented amount of mass spectrometry-based proteomics data is publicly available through repositories such as the PRoteomics IDEntifications Database (PRIDE), and the field is increasingly leveraging machine-learning approaches. However, the available data is not ready to be reused in a scalable way beyond the original acquisition purpose. Existing machine learning models commonly rely on a few manually curated datasets that require deep domain expertise and tedious technical work to construct. Importantly, these datasets have not been updated in recent years, so that newly published data remains inaccessible. We present usiGrabber, a scalable framework for assembling large proteomic datasets. usiGrabber is designed around portability and extensibility. It extracts spectra identification data from mzIdentML files, stores additional project-level metadata retrieved through the PRIDE API, indexes raw spectra using Universal Spectrum Identifiers (USIs), and offers download utilities to retrieve spectra data at scale. RESULTS: Within 49 h, we parsed over 800 million peptide spectrum matches and corresponding USIs from over 1200 projects. As a proof of concept, we used usiGrabber to construct a phosphorylation-specific training dataset of nearly 11 million spectra in under 2 days and used it to retrain a binary phosphorylation classifier based on the AHLF model architecture. With a balanced accuracy of 0.78, our model achieves comparable performance to the original model on an independent test set, showing that automated data extraction is an alternative to manual curation of static datasets. AVAILABILITY AND IMPLEMENTATION: All code is available at https://github.com/usiGrabber/usiGrabber; the data are available at https://zenodo.org/records/18853258.

Machine Learning↗

PubMed2024

Synthetic DNA barcodes identify singlets in scRNA-seq datasets and evaluate doublet algorithms.

Single-cell RNA sequencing (scRNA-seq) datasets contain true single cells, or singlets, in addition to cells that coalesce during the protocol, or doublets. Identifying singlets with high fidelity in scRNA-seq is necessary to avoid false negative and false positive discoveries. Although several methodologies have been proposed, they are typically tested on highly heterogeneous datasets and lack a priori knowledge of true singlets. Here, we leveraged datasets with synthetically introduced DNA barcodes for a hitherto unexplored application: to extract ground-truth singlets. We demonstrated the feasibility of our framework, "singletCode," to evaluate existing doublet detection methods across a range of contexts. We also leveraged our ground-truth singlets to train a proof-of-concept machine learning classifier, which outperformed other doublet detection algorithms. Our integrative framework can identify ground-truth singlets and enable robust doublet detection in non-barcoded datasets.

Algorithms↗

PubMed2026

Pretraining improves prediction of genomic datasets across species.

MOTIVATION: Recent studies suggest that deep neural network models trained on thousands of human genomic datasets can accurately predict genomic features, including gene expression and chromatin accessibility. However, training these models is computation- and time-intensive, and datasets of comparable size do not exist for most other organisms. RESULTS: Here, we identify modifications to an existing state-of-the-art model that improve model accuracy while reducing training time and computational cost. Using this streamlined model architecture, we investigate the ability of models pretrained on human genomic datasets to transfer performance to a variety of different tasks. Models pretrained on human data but fine-tuned on genomic datasets from diverse tissues and species achieved significantly higher prediction accuracy while significantly reducing training time compared to models trained from scratch, with Pearson correlation coefficients between experimental results and predictions as high as 0.8. Further, we found that including excessive training tasks decreased model performance and that this decrease could be partially but not completely rescued by fine-tuning. Thus, simplifying model architecture, applying pretrained models, and carefully considering the number of training tasks may be effective and economical techniques for building new models across data types, tissues, and species. AVAILABILITY AND IMPLEMENTATION: Code is available on GitHub and Figshare: https://github.com/optimizedlearning/genomicsML, https://doi.org/10.6084/m9.figshare.31796116.

Genomics↗

PubMed2025

MelanoDB: A dataset of clinical and molecular features of patients with advanced melanoma treated with MAPK inhibitors.

MAPK inhibitors (MAPKi) have revolutionized the treatment of patients with advanced melanoma. However, primary and acquired resistance mechanisms limit their efficacy. Predicting MAPKi response from the tumor baseline features remains challenging due to the limited size of patient cohorts. Therefore, we collected data from nine different patient cohorts (total n = 417 patients with advanced melanoma treated with MAPKi) to identify clinical and molecular features. Our curated dataset, named MelanoDB, includes whole or partial exome sequencing data for 191 patients, copy number alteration information for 66 patients, and gene expression data for 132 patients. We provide a web application to explore the integrated dataset and data distribution across the collected studies, and we share this dataset with the scientific community according to the Findable, Accessible, Interoperable, Reusable (FAIR) principles.

Humans↗

PubMed2025

De novo clustering of large long-read transcriptome datasets with isONclust3.

MOTIVATION: Long-read sequencing techniques can sequence transcripts from end to end, greatly improving our ability to study the transcription process. Although there are several well-established tools for long-read transcriptome analysis, most are reference-based. This limits the analysis of organisms without high-quality reference genomes and samples or genes with high variability (e.g. cancer samples or some gene families). In such settings, analysis using a reference-free method is favorable. The computational problem of clustering long reads by region of common origin is well-established for reference-free transcriptome analysis pipelines. Such clustering enables large datasets to be split roughly by gene family and, therefore, an independent analysis of each cluster. There exist tools for this. However, none of those tools can efficiently process the large amount of reads that are now generated by long-read sequencing technologies. RESULTS: We present isONclust3, an improved algorithm over isONclust and isONclust2, to cluster massive long-read transcriptome datasets into gene families. Like isONclust, isONclust3 represents each cluster with a set of minimizers. However, unlike other approaches, isONclust3 dynamically updates the cluster representation during clustering by adding high-confidence minimizers from new reads assigned to the cluster and employs an iterative cluster-merging step. We show that isONclust3 yields results with higher or comparable quality to state-of-the-art algorithms but is 10-100 times faster on large datasets. Also, using a 256 Gb computing node, isONclust3 was the only tool that could cluster 37 million PacBio reads, which is a typical throughput of the recent PacBio Revio sequencing machine. AVAILABILITY AND IMPLEMENTATION: https://github.com/aljpetri/isONclust3.

Algorithms↗

PubMed2026

Combining Annotation Software to Identify Orthologous Genes (CASIO) Provides a New Dataset of Orthologous Genes for Swallowtail Butterflies.

With the massive increase in genomic resources, it is becoming increasingly popular to analyse thousands of loci across many species. However, many of the available genomes are not annotated, which hinders an efficient search for orthologous protein-coding genes. Here, we aim to develop a semi-automated pipeline and compare four genomic annotation methods (BRAKER2, BUSCO, Miniprot and Scipio). Our results highlight the importance of integrating multiple annotation tools to optimise ortholog detection and improve genomic studies. Each annotation method showed different strengths. BRAKER2 annotated a substantial number of genes. BUSCO, despite limitations inherent to its reference database, identified a higher number of orthologs. Miniprot exhibited notable flexibility in accommodating diverse protein datasets, whereas Scipio successfully recovered a considerable set of genes that were not detected by the other tools. The combination of these tools allowed for more comprehensive ortholog detection. Taking advantage of this pipeline, we developed a comprehensive dataset of orthologous genes for swallowtail butterflies (Lepidoptera: Papilionidae), called Papilionidae_odb, which will facilitate future studies, especially for a non-model group with abundant genomic data and few transcriptomic resources. We tested Papilionidae_odb by inferring a robust phylogenetic framework for Leptocircini using 142 complete genomes, which improved branch support for some phylogenetic relationships, although challenges remained in resolving relationships within certain species groups, likely due to rapid radiations. Our results highlight the complementary nature of the annotation methods and suggest that combining these tools can yield more accurate results in genomic research. This approach was implemented in a Snakemake workflow called CASIO (Combining Annotation Software to Identify Orthologous genes) and can easily be applied to other non-model groups to improve genomic datasets in diverse taxa where transcriptomic resources are still limited.

Animals↗

PubMed2025

Expression regulation network in papillae of sea cucumbers: Whole-transcriptome and DNA methylation datasets.

To elucidate the expression regulation network of papilla size of sea cucumbers (Apostichopus japonicus), the whole-transcriptome and DNA methylome datasets of different sizes of papillae in sea cucumbers were generated. Average clean bases of whole-transcriptome (16.35 G) and DNA methylome (28.92 G) were obtained using RNA sequencing and whole-genome bisulfite sequencing techniques. A total of 3,188 ceRNA networks were also identified including 3,081 long non-coding RNAs (lncRNA)/microRNAs (miRNA)/mRNA networks and 107 circular RNA (circRNA)/miRNA/mRNA networks. Methylome data indicate that there were 3,307 and 3,776 differentially methylated regions (DMRs) with high-level methylation as well as 3,125 and 3,016 DMRs with low-level methylation in big papillae compared to small papillae. The identified DMRs were mainly distributed in introns, promotors, or exons. The whole-transcriptome and DNA methylome datasets generated from this study not only established a robust theoretical foundation (especially from the epigenetic aspect) for elucidating expression regulation network determining papilla size in sea cucumbers but also can be a valuable resource of biomarker mining for papilla appearance-based selective breeding in sea cucumbers.

DNA Methylation↗

PubMed2025

A dataset of estimated heterozygous individual and carrier couple frequencies for pan-ancestry carrier screening.

The data described in this publication supported the development and evaluation of pan-ancestry reproductive carrier screening panels for autosomal recessive (AR) and X-linked (XL) conditions. Raw data included combined sets of DNA variants in 1,350 AR/XL genes obtained from the ClinVar and gnomAD databases. The dataset enabled calculations of positive yield for individuals and couples across both ancestry-specific and pan-ancestry, optimised "Goldilocks"-ranked gene panels, addressing population-specific variations in the frequencies of heterozygous individuals and carrier couples. The positive yield analysis offered a performance metric for carrier screening panels, facilitating the modeling of screening performance for panels of varying sizes and composition and providing resources for optimizing panel content to ensure equity across underrepresented genetic ancestries The dataset can support ongoing research into the equitable application of carrier screening and offers significant reuse potential for refining population genetic screening practices, validating computational models, and developing frameworks to update carrier screening panels in alignment with evolving genomic data, including in underrepresented and minority populations.

Carrier screening↗

PubMed2026

Individual Differences in Cognitive Aging Rodent Datasets (ID-CARD): A collaborative platform for behavioral analysis across the lifespan.

Understanding cognitive aging requires approaches that capture individual variability while enabling integration across studies. In rodent models, behavioral data are central to this effort, yet cross-laboratory differences in experimental design limit comparability and constrain secondary analysis. To address this gap, we developed the Individual Differences in Cognitive Aging Rodent Datasets (ID-CARD), a first-of-its-kind collaborative repository aggregating trial-level Morris water maze data from multiple laboratories. ID-CARD is designed to support large-scale, integrative analyses and to facilitate secondary use of existing behavioral data in alignment with emerging data-sharing and transparency initiatives. Rather than imposing retrospective harmonization of experimental protocols, we implemented a normalization and modeling framework that enables comparison of learning trajectories while preserving meaningful variation across studies. Behavioral data from > 5000 rats spanning common strains, both sexes, and multiple ages were normalized in training and performance domains and fit with a logarithmic function to derive an error accumulation rate coefficient (EARC) as a measure of spatial learning. Age was strongly associated with increased EARC, indicating attenuated learning, even after adjusting for non-spatial cue performance. Analyses of goodness of fit revealed systematic structure in learning dynamics, where age was associated with reduced learning-curve conformity after accounting for overall performance. Inter-individual variability in spatial learning also increased with age, with strain-specific interactions. These findings demonstrate that integrated analysis of heterogeneous behavioral datasets can yield robust, individual-level insights into cognitive aging. ID-CARD provides a scalable resource and analytic framework to advance discovery in behavioral neuroscience by enabling reuse, integration, and comparative analysis of existing data.

Cognitive aging↗

PubMed2026

Rotavirus vaccine effectiveness against rotavirus and acute gastroenteritis mortality: an analysis of pooled case-control studies from the MNSSTER-V dataset.

BACKGROUND: Rotavirus accounts for an estimated 25% of diarrhoea deaths in children under 5 years globally, and more than 140 countries have included rotavirus vaccines in their routine national infant vaccination programmes. We aimed to calculate rotavirus vaccine effectiveness against rotavirus-positive and all-cause acute gastroenteritis deaths. METHODS: The Multi-National Subpopulations Study to Evaluate Rotavirus Vaccines (MNSSTER-V) dataset combines child-level data from test-negative case-control studies of rotavirus vaccine effectiveness that enrolled children under 5 years of age seeking care for acute gastroenteritis at hospitals or emergency departments in 24 countries between July 1, 2007, and Aug 24, 2023. Children were included in this study if they were: younger than 5 years, met the acute gastroenteritis case definition (had at least three episodes of diarrhoea in a 24-h period, had non-bloody and non-chronic diarrhoea, and were enrolled within 7 days of diarrhoea onset), met vaccine card quality metrics, had vaccine delivery dates if the child was reported to have received a rotavirus vaccine, and had a reported outcome of death or discharge. In-hospital acute gastroenteritis deaths were characterised, and rotavirus vaccine effectiveness against all-cause and rotavirus-positive acute gastroenteritis mortality was calculated using an unconditional logistic regression model with adjustment for national under-5 mortality strata and child's age. Vaccine effectiveness analyses against all-cause and rotavirus-positive acute gastroenteritis mortality were restricted to children aged at least 3 months who received any routine vaccines from countries reporting at least one acute gastroenteritis death. FINDINGS: From the MNSSTER-V dataset, we included 27 252 children younger than 5 years enrolled from 22 countries; outcomes of patients were not available for two countries. At least one in-hospital acute gastroenteritis death was reported from 16 countries including 21 522 children; in total, 183 all-cause acute gastroenteritis deaths and 25 rotavirus-positive deaths were reported. Among children aged at least 3 months who had received any routine vaccines, receiving at least one dose of a rotavirus vaccine had an adjusted vaccine effectiveness of 75·8% (95% CI 28·4 to 91·8; n=13 630) against rotavirus-positive acute gastroenteritis mortality and 20·8% (-47·0 to 57·3; n=20 005) against all-cause acute gastroenteritis mortality. INTERPRETATION: Rotavirus vaccines are effective in preventing rotavirus-positive acute gastroenteritis mortality. Continued efforts to improve vaccine delivery could help to reduce acute gastroenteritis mortality due to rotavirus worldwide. FUNDING: None.

Humans↗

PubMed2026

RR-interval-based atrial fibrillation detection and burden estimation: cross-dataset validation and calibration-aware probability analysis.

Objective.Atrial fibrillation (AF) burden has become an increasingly important endpoint in long-duration rhythm monitoring, but reliable burden estimation requires more than accurate AF detection alone. In particular, when burden is derived by aggregating predicted AF probabilities over time, probability calibration may directly affect burden validity under external dataset shift.Approach.This study developed an interpretable-interval feature model for AF detection and evaluated it using record-wise cross-validation on a development cohort and independent cross-dataset external validation on public Holter electrocardiographic databases. Window-level performance was assessed using the area under the receiver operating characteristic curve (ROC-AUC), area under the precision-recall curve (PR-AUC), Brier score, expected calibration error (ECE), and calibration intercept and calibration slope. Recording-level AF burden was estimated using both probability-based and hard-label aggregation and evaluated using mean absolute error (MAE) and agreement analyses.Main results.The model showed high discrimination in both development and external evaluation, with external ROC-AUC ofand PR-AUC of. However, external calibration deteriorated despite preserved ranking performance, with Brier score of, ECE(15) of, calibration intercept of, and calibration slope of. In the external cohort, probability-based burden estimation preserved strong association with reference burden but showed weaker raw agreement than hard-label aggregation, with MAE ofversus, consistent with systematic probability underprediction. Repeated external recalibration across record-level splits substantially improved probability quality and probability-based burden estimation. Median probability-burden MAE decreased fromwithout recalibration toafter Platt recalibration andafter isotonic recalibration, while median ECE(15) decreased fromtoand, respectively.Significance.These findings indicate that-interval-based AF detection maintained strong ranking performance in the tested external cohort, but probability calibration should be evaluated explicitly when predicted probabilities are aggregated into AF-burden estimates.

Atrial Fibrillation↗

PubMed2026

Features affecting Cas9-induced editing efficiency and patterns in tomato: evidence from a large CRISPR dataset.

CRISPR/Cas9 is a cornerstone of plant genome editing, yet the determinants of editing efficiency for a given single-guide RNAs (sgRNAs) and DNA double-strand break (DSB) repair outcomes remain poorly understood, particularly in plants. Here, we generated a large experimental dataset comprising 420 sgRNAs targeting promoters, exons, and introns of 137 genes in tomato protoplasts, and quantified editing efficiency and repair footprints together with chromatin accessibility and transcriptional state in the same cellular context. Editing efficiency was consistently higher at targets in accessible chromatin and modestly higher in promoters and introns than in exons, whereas transcriptional activity had no detectable effect. Editing efficiencies were more similar among sgRNAs targeting the same gene than among different genes, revealing a local genomic influence on Cas9 activity. A distinct subset of sgRNAs achieved near-complete editing and produced characteristic repair footprints dominated by long deletions with extended microhomology tracts, indicative of microhomology-mediated end joining (MMEJ), resembling patterns associated with high-efficiency guides in human cells, and suggesting conserved sequence-driven repair biases across species. In contrast, widely used human-trained prediction models failed to accurately rank sgRNA performance in plants, highlighting the limits of cross-species predictability. Together, this dataset provides a resource for improving guide design and mechanistic understanding of plant DNA repair.

Solanum lycopersicum↗

PubMed2026

UnionLoops: a workflow for calling chromatin loops across related Hi-C datasets with improved specificity, precision, and sensitivity.

Chromatin loop calling from chromatin interaction data often exhibits substantial variability across related samples. We present UnionLoops, a computational workflow for chromatin loop calling across multiple related samples. UnionLoops integrates information across datasets to determine positions and dataset-specificity of looping interactions. It constructs a unified candidate loop set, applies consistent filtering and aggregation, and evaluates loop support across samples. We demonstrate that UnionLoops increases sensitivity for detecting shared chromatin loops, reduces spurious sample-specific calls, and improves concordance with independent genomic features, including CTCF and cohesin occupancy. UnionLoops enables improved biological interpretation of chromatin loop organization and dynamics across related conditions.

Chromatin↗

PubMed2025

Viral genome sequence datasets display pervasive evidence of strand-specific substitution biases that are best described using non-reversible nucleotide substitution models.

Most phylogenetic trees are inferred using time-reversible evolutionary models that assume that the relative rates of substitution for any given pair of nucleotides are the same regardless of the direction of the substitutions. However, there is no reason to assume that the underlying biochemical mutational processes that cause substitutions are similarly symmetrical. We consider two non-reversible nucleotide substitution models: (1) a 6-rate non-reversible model (NREV6) that is applicable to analyzing mutational processes in double-stranded genomes in that complementary substitutions occur at identical rates; and (2) a 12-rate non-reversible model (NREV12) that is applicable to analyzing mutational processes in single-stranded (ss) genomes in that all substitution types are free to occur at different rates. Using likelihood ratio and Akaike Information Criterion-based model tests, we show that, surprisingly, NREV12 provided a significantly better fit than the General Time Reversible (GTR) and NREV6 models to 21/31 dsRNA and 20/30 dsDNA datasets. As expected, however, NREV12 provided a significantly better fit to 24/33 ssDNA and 40/47 ssRNA datasets. We tested how non-reversibility impacts the accuracy with which phylogenetic trees are inferred. As simulated degrees of non-reversibility (DNR) increased, the tree topology inferences using both NREV12 and GTR became more accurate, whereas inferred tree branch lengths became less accurate. We conclude that while non-reversible models should be helpful in the analysis of mutational processes in most virus species, there is no pressing need to use these models for routine phylogenetic inference.

Models of evolution↗