Biology subjects

Nekrutenko, A.

Publications and source records attributed to Nekrutenko, A..

At least 19 recordsLinked to original sources

bioRxiv2026

Verifiable Agentic Science on Public Infrastructure

Autonomous artificial intelligence agents are transforming the practice of science. Across every discipline, researchers are handing data analysis to coding agents that can design and execute complex studies from a single prompt. Yet this revolution carries a threat. When agents improvise in local environments, they leave no verifiable record, ultimately risking a wave of irreproducible science. Solving this crisis requires moving agentic science onto public infrastructure where every action leaves an examinable trail. Here we connect autonomous coding agents to shared computing resources, turning ephemeral prompts into auditable scientific records. After showing that 30 unguided local agents splinter into 18 incompatible pipelines on the same dataset, we conduct an agentic reproduction and discovery study on a large scale---reproducing a large, diverse array of studies with every workflow, history, and intermediate dataset publicly accessible. This complete provenance trail exposes unstated modeling choices, reference mismatches, and subsampled tutorial inputs in published studies, and establishes a new foundation where human scientists can understand, steer, and trust AI-driven discovery.

bioinformatics↗

bioRxiv2026

Clean Breaks: Attention Force Field Deconvolution of Recombination Across Sequence Manifolds

Homologous recombination and horizontal gene transfer assemble mosaic chromosomes across divergent lineages. Reticulate evolution violates bifurcating trees, yet existing detectors enforce a stark trade-off: phylogenetic changepoint models optimize partitioned topologies at exponential computational cost, whereas sequence triplet heuristics evaluate rapidly but suffer from severe combinatorial event overcounting. We introduce RhizAeon, an open-source platform that deconvolves reticulate evolution through continuous trajectory mechanics across metric embeddings in linear time (O(N K L)). Projecting aligned sequences into canonical metric coordinates and routing lineage-specific rate variation into an unpolarized root sink, the engine iteratively scrubs mosaic tracts and verifies breakpoints through Brownian-bridge statistics, dyadic matched filters, and maximum-likelihood polishing. Across canonical coalescent benchmarks and a continuous parameter sweep of 1,000,000 whole genomes, RhizAeon bounds false-positive rates well below the 5.0% nominal ceiling (0.58% on 1,200 rate-variation nulls and 1.01% across 333,334 continuous sweep nulls). Leave-one-in verification and bidirectional overlap clustering collapse co-recombinant clades into single ancestral reticulations. Empirical analyses across wildlife sarbecovirus Spike chimeras, multidrug-resistant Streptococcus pneumoniae chromosomes (2.22 Mb), and 379 primate HLA Class I alleles reconstruct complex whole-genome mosaicism, cassette exchanges, and micro-conversions in seconds. The core engine compiles to WebAssembly for interactive execution within standard web browsers.

evolutionary biology↗

bioRxiv2026

PyEuk: a tool suite for catalogue-free multilocus typing

Multilocus sequence typing anchors molecular epidemiology, but traditional frameworks require centrally curated allele catalogues. For emerging and uncultivable eukaryotic parasites, maintaining these databases is impractical, leaving surveillance reliant on fragmented, assay-specific scripts. PyEuk eliminates this bottleneck by providing an open, catalogue-free suite that calls microhaplotypes directly from sequence differences relative to a reference within data-defined genomic windows. When amplicon coordinates are uncharacterized or unpublished, PyEuk reconstructs target panels de novo from raw read coverage peaks mapped to a draft assembly. Across benchmark cohorts spanning Cyclospora cayetanensis and Plasmodium vivax, PyEuk recovers epidemiological structure established by tracebacks, geography, and clinical recurrence without organism-specific tuning. In foodborne outbreaks, it resolves independent transmission chains using either curated or de novo panels and scales to national surveillance archives exceeding 8,000 isolates. In P. vivax malaria, its weighted identity-by-state distance separates continental lineages, discriminates liver-stage relapses from reinfections, and delineates transmission clusters. Rather than forcing an arbitrary partition on continuous variation, PyEuk evaluates bootstrap stability, reporting supported cluster count ranges alongside reproducible transmission cores. PyEuk provides a portable, reproducible foundation for eukaryotic pathogen surveillance.

genomics↗

bioRxiv2026

HyphAeon: Attention on Evolution Across Deep Time Transforms Comparative Genomics

Detecting Darwinian natural selection is fundamental to evolutionary biology and functional genomics, yet standard methods based on phylogenetic models that estimate the ratio of non-synonymous to synonymous substitution rates (dN/dS) fail to scale with modern genomic volumes. Fitting continuous-time Markov substitution matrices across dense trees with hundreds of species requires extensive compute, forcing comparative genomics to rely on aggressive taxon subsampling or static whole-tree summaries that dilute transient adaptive bursts. Here we present HyphAeon, a lightweight (~1.91M parameter backbone, 2.46M across the full multi-task suite) phylogeny-informed foundation transformer trained to amortize the detection of episodic diversifying selection across 742-species mammalian coding alignments (17,186 genes, 9.77 x 10^6 codons). HyphAeon approaches the discriminative accuracy of numerical maximum-likelihood selection tests (MEME) across episodic burst regimes (ROC-AUC up to 0.942, mean 0.659; empirical Precision-Recall lift up to 25.8x, mean 7.7x; rank concordance up to rho = 0.983) while executing >1,000x faster per locus (averaging <1 ms per site across genome-wide scans) and >10,000x faster at proteome scale, generalizing outside its mammalian training distribution without retraining. Beyond accelerating classical tests, embedding molecular evolution into a differentiable geometric latent space enables analytical capabilities inaccessible to static dN/dS models: (1) targeted alignment artifact correction via counterfactual attribution; (2) macromolecular contact recovery and multi-site epistatic sectors (CESI); (3) directional phenotype-to-genotype attribution in lineage space (PARS); and (4) continuous temporal surveillance regression that tracks positive sweep velocities across longitudinal cohorts (evaluated across 12,167 timestamped genomes and benchmarked against external frequencies from >9.34 million genomes), rescuing adaptive substitutions obscured by post-fixation dilution. By bridging statistical phylogenetics with geometric representation learning, HyphAeon establishes comparative genomics as an interactive, high-throughput computational framework for evolutionary discovery.

bioinformatics↗

bioRxiv2026

Evaluating open LLMs for agentic analysis orchestration in a typical biomedical lab

Agentic tools -- software environments where a large language model plans, calls external tools, executes code, and iterates with minimal human intervention -- will run a substantial share of routine biomedical data analysis within the next few years. However, per-call inference cost on frontier models is the bottleneck and can add up quickly. Here, we tested whether a free, locally-runnable open-weight model could take over the repetitive execution steps at frontier accuracy. We used Claudes Opus to author plans of increasing detail for per-sample variant calling, and ran six 2026-release open-weight implementer LLMs against those plans on a set of desktop GPUs. qwen3.6:27b reproduced frontier accuracy on every plan and matched Opus cell-for-cell on a 36-cell error-injection matrix. A sub-$2,000 Jetson or Apple Mac Mini sufficed for the implementer side. The open-weight model landscape evolves on the order of months, so the specific implementer recommended here will be superseded; we provide the plans, harness, scoring code, and per-cell artifacts at https://github.com/nekrut/LLM-eval-paper as a framework for re-evaluating future models.

bioinformatics↗

bioRxiv2026

The Common Fund Data Ecosystem (CFDE)

The NIH Common Fund Data Ecosystem (CFDE) integrates data resources from 18 NIH Common Fund programs for discovery and integrative analysis. These programs generate valuable but heterogeneous datasets that can be difficult to discover, access, and reuse. CFDE aims to provide a collaborative, community-built infrastructure that links and enriches Common Fund programs. We describe the evolution, structure, and core technologies of CFDE, including practical approaches that support submission, integration, visualization, and public release of multimodal data. Training programs and workforce initiatives lower barriers to adoption. CFDE has devised solutions to critical issues facing cross-program initiatives, including data scale and heterogeneity, dataset integration, and long-term sustainability. We demonstrate the utility of linking Common Fund resources through integrative tools and cross-dataset queries to yield insights that would otherwise be infeasible. Collectively, CFDE shows that a standards-driven, federated approach enhances and unifies cross-disciplinary resources, fostering collaboration and data-driven discovery.

scientific communication and education↗

bioRxiv2026

Beyond Invariable Sites: Using Evolutionary Stasis to Map Multi-Layered Constraints on the Evolution of Viral and Mammalian Genomes

The quantification of genomic conservation has progressed from foundational statistical modeling of evolutionary rates to state-of-the-art phylogeny-aware deep learning architectures. Yet, a fundamental resolution gap remains whenever evolutionary rates closely approach the "zero-rate origin," where standard selection inference tools will essentially ignore signals of extreme purifying section at invariant genome sites. We present B-STILL (Bayesian Significance Test of Invariant Low Likelihoods), a hierarchical Bayesian framework designed to resolve the selective landscape of protein-coding data by leveraging gene-level calibration and codon-site specific evolutionary opportunity. This framework is based on computationally efficient approximations using codon-substitution models which are scalable to alignments with thousands of sequences. By explicitly tuning the stasis radius around the near-zero evolutionary-rate regime, B-STILL distinguishes between stochastic invariance and functional constraint, identifying Evolutionary Stasis Anchors (ESAs) where the upper bound on permitted evolutionary change is statistically anomalous relative to the background of the gene. This hierarchical approach provides a signature of functional or structural constraint that is often difficult to detect using other tools. Validation against extensive pathogen and clinical databases confirms that ESAs are predictors of biological fitness and disease potential. Collectively, we identified thousands of significantly clustered ESAs that precisely footprint both known functional domains and currently uncharacterized structural motifs in mammalian and viral genomes. These findings establish B-STILL as a scalable statistical framework for high-resolution genomic annotation, transforming formerly ignored invariant genome and protein sites into informative markers of extreme purifying selection across both well-characterized and uncharacterized protein-coding genes from different domains of life.

evolutionary biology↗

bioRxiv2026

Characterizing Physicochemical Selection in Protein Evolution with Property-Informed Models (PRIME)

AbstractStandard probabilistic models of coding sequence evolution effectively identify where and when selection acts but remain agnostic to the mechanistic realization of these forces. We introduce PRIME (PRoperty Informed Models of Evolution), a framework of codon-level maximum likelihood methods--including global (G-PRIME), episodic (E-PRIME), and site-specific (S-PRIME) implementations--that explicitly model amino acid exchange-ability as a function of physicochemical properties. By parameterizing attributes such as molecular volume, hydropathy, and secondary structure propensities, PRIME resolves the biophysical basis of selective constraint across both the sequence and the phylogeny. At the site level, S-PRIME leverages an explicit biophysical taxon-omy to precisely categorize residues as conserved, neutral, or changing for specific properties, resolving selective signals that remain invisible to traditional rate-based metrics. Our analysis of a benchmark of 24 diverse datasets and a genome-wide screen of 18,944 mammalian genes demonstrates that biophysical realism yields substantial improvements in model fit, acting synergistically with rate variation to explain complex evolutionary patterns. We find that power to detect physicochemical constraints at individual sites is fundamentally governed by simple informational redundancy (substitutions per unique amino acid; AUC = 0.91), with sensitivity exceeding 90% in data-rich alignments. E-PRIME reveals a distinct biophysical hierarchy: while core packing and beta-sheet scaffolds are rigidly conserved, alpha-helix propensity and surface electrostatics serve as the primary substrates for adaptive tuning. Furthermore, PRIME importance weights align with aspects of the primary semantic axes of deep learning representations (ESM-2) and capture key features of experimental fitness landscapes. By trans-forming abstract evolutionary rates into interpretable biophysical rules, PRIME provides a useful framework for characterizing the mechanistic drivers of protein diversity.

evolutionary biology↗

bioRxiv2026

BUSTED-PH: Isolating the genomic signatures of convergent phenotypes

Convergent evolution offers a natural test for adaptive predictability, yet pinpointing its molecular basis remains difficult. Current methods often grapple with distinguishing adaptive convergence from background noise, either by demanding overly stringent identical substitutions or relying on low-resolution evolutionary rate shifts. Here we introduce BUSTED-PH (Branch-site Unrestricted Statistical Test for Episodic Diversification - Phenotype), a branch-site codon model that detects phenotype-associated episodic diversifying selection. Unlike standard approaches, BUSTED-PH explicitly contrasts selective regimes between phenotype-positive (foreground) and phenotype-negative (background) lineages, effectively winnowing out spurious associations driven by pervasive background adaptation. BUSTED-PH has already been applied in numerous independent studies, and here we rigorously validate it using canonical positive controls and simulations to confirm high power and strict false positive control. A genome-wide scan of 120 mammalian species using strict statistical criteria (FDR [≤] 0.01) identified 72 genes associated with echolocation; while recovering paradigmatic auditory drivers (e.g., Prestin, TMC1 ), we also uncover novel candidates in physiological support systems ranging from lipid homeostasis to neural development. A parallel analysis of mammalian gigantism identifies 91 genes linked to musculoskeletal reinforcement, organ size governance, and genomic integrity, characterizing the molecular adaptations required to support massive body size. As projects like Zoonomia and the Vertebrate Genomes Project expand comparative datasets, BUSTED-PH provides a robust framework for dissecting the genetic architecture of complex convergent traits.

evolutionary biology↗

bioRxiv2025

Standardizing RNA-seq Analysis of Fungal Pathogens Using BRC-Analytics and Agentic AI: A Candidozyma auris Case Study

Candidozyma auris has emerged as a critical global health threat due to multidrug resistance and healthcare-associated transmission. While RNA-seq has become the primary tool for studying C. auris pathogenesis, inconsistent use of reference genomes and bioinformatics tools complicate cross-study comparisons. Here we demonstrate how BRC-Analytics, a platform for pathogen genomics, combined with an agentic AI assistant, enables reproducible RNA-seq analysis. By re-analyzing data from two publications we achieved near-perfect correlation with published results despite annotation version differences. We addressed provenance challenges associated with using AI agents with Galaxy by forcing them to invoke Galaxys native tools rather than manipulating data directly. For custom analyses outside Galaxys toolset, we provide standalone JupyterLite notebooks that reproduce our analysis without AI involvement. This framework--combining AI-assisted automation with rigorous provenance tracking--establishes a template for standardized, reproducible fungal pathogen genomics. To the best of our knowledge, this is the first example of integration between public data repositories, reproducible analysis workflows, and agentic AI tools. Our subsequent efforts will focus on improving the seamlessness of this integration.

genomics↗

bioRxiv2025

From data to publication in a browser with BRC-Analytics: Evolutionary dynamics of coding overlaps in measles virus

The analytical landscape of pathogen research is often fragmented, hindering transparency and reproducibility due to diverse genomic data sources, numerous software tools, and suboptimal integration methods. Here we introduce BRC-analytics, a novel browser-based environment that unifies authoritative sources of genomic data with community-curated best analysis practices on a freely accessible public computational infrastructure. We demonstrate its capabilities by analyzing the evolutionary dynamics within the P/V/C locus of the measles virus, a complex system involving overlapping coding regions and RNA editing. Our analysis, conducted entirely within BRC-analytics, reveals asymmetric evolution of the locuss reading frames under distinct selective pressures. BRC-analytics streamlines the entire research process--from data collection and primary analysis (e.g., variant calling) to interpretation (e.g., using integrated JupyterLite notebooks and LLMs) and publication--into a single web browser session. This eliminates the need for local installations and manual data transfers, implicitly tracking provenance and ensuring reproducibility. The platforms goal is to provide true data-to-publication functionality, making advanced pathogen genomics accessible to a broader research community regardless of their computational expertise or infrastructure access.

genomics↗

bioRxiv2024

Minus the Error: Estimating dN/dS and Testing for Natural Selection in the Presence of Residual Alignment Errors

Positive selection is an evolutionary process which increases the frequency of advantageous mutations because they confer a fitness benefit. Inferring the past action of positive selection on protein-coding sequences is fundamental for deciphering phenotypic diversity and the emergence of novel traits. With the advent of genome-wide comparative genomic datasets, researchers can analyze selection not only at the level of individual genes but also globally, delivering systems-level insights into evolutionary dynamics. However, genome-scale datasets are generated with automated pipelines and imperfect curation that does not eliminate all sequencing, annotation, and alignment errors. Positive selection inference methods are highly sensitive to such errors. We present BUSTED-E: a method designed to detect positive selection for amino acid diversification while concurrently identifying some alignment errors. This method builds on the flexible branch-site random effects model (BUSTED) for fitting distributions of dN/dS, with a critical modification: it incorporates an "error-sink" component to represent an abiological evolutionary regime. Using several genome-scale biological datasets that were extensively filtered using state-of-the art automated alignment tools, we show that BUSTED-E identifies pervasive residual alignment errors, produces more realistic estimates of positive selection, reduces bias, and improves biological interpretation. The BUSTED-E model promises to be a more stringent filter to identify positive selection in genome-wide contexts, thus enabling further characterization and validation of the most biologically relevant cases.

evolutionary biology↗

bioRxiv2024

KegAlign: Optimizing pairwise alignments with diagonal partitioning

Our ability to generate sequencing data and assemble it into high quality complete genomes has rapidly advanced in recent years. These data promise to advance our understanding of organismal biology and answer longstanding evolutionary questions. Multiple genome alignment is a key tool in this quest. It is also the area which is lagging: today we can generate genomes faster than we can construct and update multiple alignments containing them. The bottleneck is in considerable computational time required to generate accurate pairwise alignments between divergent genomes, an unavoidable precursor to multiple alignments. This step is typically performed with lastZ, a very sensitive and yet equally slow tool. Here we describe an optimized GPU-enabled pairwise aligner KegAlign. It incorporates a new parallelization strategy, diagonal partitioning, with the latest features of modern GPUs. With KegAlign a typical human/mouse alignment can be computed in under 6 hours on a machine containing a single NVidia A100 GPU and 80 CPU cores without the need for any pre-partitioning of input sequences: a [~]150x improvement over lastZ. While other pairwise aligners can complete this task in a fraction of that time, none achieves the sensitivity of KegAligns main alignment engine, lastZ, and thus may not be suitable for comparing divergent genomes. In addition to providing the source code and a Conda package for KegAlign we also provide a Galaxy workflow that can be readily used by anyone.

genomics↗

bioRxiv2023

Scalable, accessible, and reproducible reference genome assembly and evaluation in Galaxy

Improvements in genome sequencing and assembly are enabling high-quality reference genomes for all species. However, the assembly process is still laborious, computationally and technically demanding, lacks standards for reproducibility, and is not readily scalable. Here we present the latest Vertebrate Genomes Project assembly pipeline and demonstrate that it delivers high-quality reference genomes at scale across a set of vertebrate species arising over the last [~]500 million years. The pipeline is versatile and combines PacBio HiFi long-reads and Hi-C-based haplotype phasing in a new graph-based paradigm. Standardized quality control is performed automatically to troubleshoot assembly issues and assess biological complexities. We make the pipeline freely accessible through Galaxy, accommodating researchers even without local computational resources and enhanced reproducibility by democratizing the training and assembly process. We demonstrate the flexibility and reliability of the pipeline by assembling reference genomes for 51 vertebrate species from major taxonomic groups (fish, amphibians, reptiles, birds, and mammals).

bioinformatics↗

bioRxiv2022

Lineage tracing via associative chromosome/plasmid barcoding with siBar

We describe a new method, siBar, which enables simultaneous tracking of plasmids and chromosomal lineages within a bacterial host. siBar involves integration of a linearized plasmid construct carrying a unique combination of two molecular barcodes. Upon recircularization one barcode remains integrated into the host chromosome, while the other remains on the plasmid allowing direct observation of expansion and contraction of adaptive lineages. We also performed a pilot evolution experiment that allowed us to assess the barcode complexity and establish an analytical framework for the analysis of siBar data.

microbiology↗

bioRxiv2022

Planemo: a command-line toolkit for developing, deploying, and executing scientific data analyses

There are thousands of well-maintained high-quality open-source software utilities for all aspects of scientific data analysis. For over a decade, the Galaxy Project has been providing computational infrastructure and a unified user interface for these tools to make them accessible to a wide range of researchers. In order to streamline the process of integrating tools and constructing workflows as much as possible, we have developed Planemo, a software development kit for tool and workflow developers and Galaxy power users. Here we outline Planemos implementation and describe its broad range of functionality for designing, testing and executing Galaxy tools, workflows and training material. In addition, we discuss the philosophy underlying Galaxy tool and workflow development, and how Planemo encourages the use of development best practices, such as test-driven development, by its users, including those who are not professional software developers. Planemo is a mature project widely used within the Galaxy community which has been downloaded over 80,000 times.

bioinformatics↗

bioRxiv2022

Selection analysis identifies unusual clustered mutational changes in Omicron lineage BA.1 that likely impact Spike function

Among the 30 non-synonymous nucleotide substitutions in the Omicron S-gene are 13 that have only rarely been seen in other SARS-CoV-2 sequences. These mutations cluster within three functionally important regions of the S-gene at sites that will likely impact (i) interactions between subunits of the Spike trimer and the predisposition of subunits to shift from down to up configurations, (ii) interactions of Spike with ACE2 receptors, and (iii) the priming of Spike for membrane fusion. We show here that, based on both the rarity of these 13 mutations in intrapatient sequencing reads and patterns of selection at the codon sites where the mutations occur in SARS-CoV-2 and related sarbecoviruses, prior to the emergence of Omicron the mutations would have been predicted to decrease the fitness of any genomes within which they occurred. We further propose that the mutations in each of the three clusters therefore cooperatively interact to both mitigate their individual fitness costs, and adaptively alter the function of Spike. Given the evident epidemic growth advantages of Omicron over all previously known SARS-CoV-2 lineages, it is crucial to determine both how such complex and highly adaptive mutation constellations were assembled within the Omicron S-gene, and why, despite unprecedented global genomic surveillance efforts, the early stages of this assembly process went completely undetected.

evolutionary biology↗

bioRxiv2022

RASCL: Rapid Assessment Of SARS-CoV-2 Clades Through Molecular Sequence Analysis

An important component of efforts to manage the ongoing COVID19 pandemic is the Rapid Assessment of how natural selection contributes to the emergence and proliferation of potentially dangerous SARS-CoV-2 lineages and CLades (RASCL). The RASCL pipeline enables continuous comparative phylogenetics-based selection analyses of rapidly growing clade-focused genome surveillance datasets, such as those produced following the initial detection of potentially dangerous variants. From such datasets RASCL automatically generates down-sampled codon alignments of individual genes/ORFs containing contextualizing background reference sequences, analyzes these with a battery of selection tests, and outputs results as both machine readable JSON files, and interactive notebook-based visualizations. AvailabilityRASCL is available from a dedicated repository at https://github.com/veg/RASCL and as a Galaxy workflow https://usegalaxy.eu/u/hyphy/w/rascl. Existing clade/variant analysis results are available here: https://observablehq.com/@aglucaci/rascl. ContactDr. Sergei L Kosakovsky Pond (spond@temple.edu). Supplementary informationN/A

bioinformatics↗