Biology subjects

Georgakopoulos-Soares, I.

Publications and source records attributed to Georgakopoulos-Soares, I..

At least 19 recordsLinked to original sources

bioRxiv2026

Arena3D 3.0: a web server for exploring, analyzing, annotating and sharing multilayered networks in 3D

Multilayer networks describe biological entities linked by several kinds of relation or measured under several conditions. Arena3D 3.0, a complete reimplementation of Arena3Dweb, gives each layer its own place in an interactive 3D scene. Networks are uploaded or imported from STRING, IntAct, OmniPath, NDEx, Reactome and the Gene Ontology. Twelve layouts, seven in 3D, are available, and node copies can be aligned across layers. Layers are drawn as floors or 3D containers in thirteen arrangements, including 3D hive plots. Communities and annotated groups can be laid out apart and outlined as hulls or fog clouds, and edges can be bundled. Analyses cover node and multilayer metrics, distributions and perturbation, with chord, Sankey, heatmap, UpSet and UMAP views. Networks go to Cytoscape, Gephi Lite, NORMA and JBrowse 2, gene lists to Flame, and a REST API gives programmatic access. Arena3D 3.0 is free at www.arena3d.org without login.

bioinformatics↗

bioRxiv2026

Calibration-free compression brings Evo 2 to its full million-token context on a single GPU

Evo 2 is the largest openly available genomic foundation model, but its forty billion parameter configuration cannot be loaded onto a single 80 GB accelerator, placing genome-scale analysis beyond most laboratories. We present TurboQuant-Bio, an open toolkit that compresses Evo 2s weights and attention cache to four bits without calibration data, and serves both through fused kernels. Compression is near-lossless across perplexity spanning the tree of life, genomic classification, splice-site prediction, gene completion and clinically relevant variant-effect prediction. It brings Evo 2 40B onto one 80 GB GPU and Evo 2 7B to its full million-token context within a 40 GB memory budget, an eightfold gain in reachable context. We further show that the released chunked-prefill path is silently incorrect, returning plausible but uncorrelated likelihoods, and derive the block-wise continuation that repairs it: a complete 580-kilobase bacterial genome is now scored in one context in 22 minutes rather than 13.7 hours.

bioinformatics↗

bioRxiv2026

ResLit: A Large-Scale Automated Literature Mining Database for Antimicrobial Resistance

Antimicrobial resistance generates a vast, rapidly growing literature, yet no resource offers a comprehensive, evidence-linked repository of AMR findings at scale. We present ResLit, an automated pipeline and public database that mines the AMR literature for resistance genes, mutations, organisms, and mechanisms. From 2 million candidate PubMed records, BioMistral-7B screened abstracts to 356,000 relevant papers; multi-tier retrieval yielded 117,000 full texts, from which Qwen3-30B performed two-step extraction. ResLit contains 3,120 genes and 13,593 mutations, cross-linked to CARD, ResFinder, and NCBI Reference Gene Catalog across four evidence tiers. It further supports community-driven curation of automated outputs and reference databases. Freely available at www.reslit.info.

microbiology↗

bioRxiv2026

Comprehensive landscape of potential CTCF-binding sites across the complete telomere-to-telomere human genome

CCCTC-binding factor (CTCF) is a master regulator of the human genome, directly modulating both transcription and 3D genome organization. However, previous human genome-wide maps of its binding sites have been built on an incomplete assembly, leaving out many repetitive, difficult-to-sequence regions. Here, we leverage the telomere-to-telomere human assembly to generate a complete catalog of potential CTCF-binding sites and characterize their function in centromeric satellites, rDNA arrays, and acrocentric short arms. Using GC-stratified permutation testing, we find that CTCF motifs are significantly enriched in the alpha-satellite active higher-order repeats of chromosomes 18, 20, and X, in rDNA arrays, and in gamma satellites. Within acrocentric short arms, CTCF motifs concentrate in the hypomethylated and accessible regions of rDNA arrays and composite beta satellites. Although short arms share high sequence homology, most potential CTCF sites are arm-specific. Cross-arm shared sites were 1.7-fold more likely to fall in the frequently recombining pseudo-homologous regions, with density scaling with sequence mosaicism. Extending our analysis to non-human primate assemblies, we show that CTCF enrichment in gamma satellites and rDNA arrays is consistent across species. Subterminal heterochromatin spacer regions contain CTCF sites deriving largely from ancient LINE elements and lineage-specific segmental duplications act as a site transfer mechanism between species. We believe that our findings will provide the foundation for future studies of CTCFs function in the repetitive parts of the human genome.

genomics↗

bioRxiv2026

Genome-wide mapping of persistent long-range correlation in the complete telomere-to-telomere human reference genome

Long-range correlations (LRCs) in DNA sequences have been reported for decades, but their interpretation has been limited by incomplete representation of repetitive and structurally complex regions in earlier human genome assemblies. Here, we use the complete telomere-to-telomere (T2T) human reference genome to systematically map LRC structure across all human chromosomes. Most chromosomes exhibited persistent long-range dependence (LRD), with chromosome-specific scaling intervals spanning kilobase to megabase scales and wavelet Hurst exponents consistently above the uncorrelated expectation of 0.5. Spatially resolved Hurst landscapes revealed that this signal is not uniformly distributed along chromosomes, but instead reflects a heterogeneous genomic mosaic. For the purine-pyrimidine encoding, chromosomes 9, 15, X, and Y showed complex fluctuation profiles not adequately summarized by a single scaling exponent; multifractal detrended fluctuation analysis (MF-DFA) confirmed broad, q-dependent multiscaling consistent with multifractal behavior in these chromosomes, consistent with heterogeneous contributions from satellite-rich and structurally distinct sequence compartments. Surrogate analyses further showed that the observed correlations exceed expectations from base composition alone and are not fully explained by local sequence structure preserved in block-shuffled controls. However, short-memory Autoregressive Moving-Average (ARMA) surrogates reproduced the observed H range in a subset of chromosomes, indicating that the strength of evidence for LRD is chromosome-dependent. Together, these results demonstrate that LRC is a widespread but spatially heterogeneous property of the complete human genome. The T2T assembly reveals that previously unresolved repetitive and satellite-rich regions are strongly associated with chromosome-scale and local scaling variation, providing a framework for future studies of genome evolution, chromatin structure, recombination, structural variation, and genomic instability.

genomics↗

bioRxiv2026

HSeeker: an algorithm for systematic H-DNA sequence identification

H-DNA is a naturally occurring intramolecular DNA triplex structure formed by Hoogsteen hydrogen bonds at homopurine-homopyrimidine mirror repeats and has functional roles in gene regulation, genome instability, and human disease. The existing H-DNA detection tools capture only a subset of H-DNA sequences, often missing relevant sequence features or failing to assess key aspects of structural stability. To address this gap, we present "HSeeker", a state-of-the-art computational tool that compiles a three-part algorithm to identify and score potential H-DNA-forming sequences. Using a center-outward search algorithm approach, HSeeker evaluates candidate hinge positions and spacer lengths while allowing configurable mirror mismatches and purine-pyrimidine composition thresholds. The greedy overlap removal phase resolves overlapping candidates by retaining the longest and most compact motif within each overlapping region. Finally, the thermodynamic stability scoring algorithm evaluates the candidate motifs using an experimentally informed scoring model that incorporates Hoogsteen G-G and A-A bonds, consecutive-pair stacking, and imposes penalties for mismatches and disrupted stacks. The scoring procedure also optimizes motif boundaries by trimming weak terminal positions and reassigning unstable arm positions to the spacer. HSeeker reports genomic coordinates, sequence information, pairing and stacking components, and an overall stability score. HSeeker is also user-friendly, available as a Python package and as a web application, supporting configurable, high-throughput analysis and exportability of predicted H-DNA motifs. HSeeker provides an accessible and reproducible framework for investigating the distribution and potential stability of H-DNA-forming sequences across genomic datasets.

genomics↗

bioRxiv2026

ArthroVerse: mapping protein family diversity across arthropod-associated microbiomes

Metagenomic studies of arthropod-associated microbiomes have generated vast amounts of sequence data, yet the functional and structural organization of these proteins remains largely unexplored. Here, we present ArthroVerse, the first comprehensive database of protein families derived from arthropod-associated metagenomes. Non-redundant protein families were generated after rigorous filtering, deduplication, and clustering. The protein families were further annotated with microbial taxonomy, host associations, protein structural information, and Carbohydrate-active enzymes (CAZyme) predictions. The resulting dataset integrates both metagenomic and reference genome-derived proteins, enabling systematic exploration of functional diversity, evolutionary relationships, and host-microbe interactions in insect microbiomes. ArthroVerse provides a valuable resource for the study of microbial ecology and arthropod physiology, offering unprecedented insight into the protein landscape of insect-associated microbial communities.

bioinformatics↗

bioRxiv2026

Safeguarding open-weight genomic foundation models through weight locking

BackgroundGenomic foundation models can dramatically accelerate biological research by learning general-purpose representations of genomic data that transfer across tasks, enabling researchers to predict variant effects, regulatory elements, and molecular function, among others. To safeguard against potential biosecurity threats and malicious misuse of open-weight models, a common strategy involves excluding human-infecting viral genomes from the models training corpora. This strategy, however, can be easily circumvented by fine-tuning models on abundantly available viral data. Weight-locking with spectral deformation has been proposed as a potential method to prevent fine-tuning of neural networks, but has not been systematically evaluated in biological AI models. MethodsWe applied spectral deformation locking to the Evo-1-8k-base genomic foundation model and evaluated a panel of attack configurations spanning naive fine-tuning, low-rank adaptation (LoRA), a simple inserted-layer bypass baseline, and a white-box singular value decomposition (SVD)-chain factorisation at chain lengths k [isin] {2, 3, 5}. Recovered virological capability was quantified on three Human Virome Understanding Evaluation (HVUE) tasks. ResultsThe lock defended against the naive attacker by either standard pipeline. Naive full fine-tuning under the strong lock drove downstream virological capability significantly below the pretrained baseline on pathogenicity and host tropism, converting the attack into a capability loss rather than a gain, while naive low-rank adaptation neither moved held-out perplexity (PPL) nor recovered downstream capability above pretrained. Thus, we conclude that by neither route does the naive attacker reach the gain achieved by fine-tuning an unlocked model. Consistent with previous results in non-biological models, an informed attacker who implements the SVD-chain construction does recover capability on pathogenicity prediction, at the cost of increased computational requirements for the fine-tuning process. Availabilityhttps://github.com/Georgakopoulos-Soares-lab/glm-locking.

bioinformatics↗

bioRxiv2026

WasteFams: A database of protein families from global wastewater microbiomes

Wastewater surveillance has emerged as a critical tool for global epidemiology, yet the functional diversity of wastewater microbiomes remains poorly characterized at the protein level. Here, we present WasteFams, the first comprehensive database dedicated to the systematic exploration of protein families in wastewater metagenomic and metatranscriptomic studies worldwide. Integrating data from 580 metagenomes, 132 metatranscriptomes, and 1,709 reference genomes, WasteFams catalogs 3,887 non-redundant protein families (containing {succeq}100 members) derived from over 105 million predicted proteins. Each protein family is enriched with multi-layered annotations, including AlphaFold3 structural predictions, taxonomic classifications, and biome-specific metadata. To further expand their functional annotation, we integrated deep genomic context analysis to link protein families to Mobile Genetic Elements (MGEs), Biosynthetic Gene Clusters (BGCs), Antibiotic Resistance Genes (ARGs), and CRISPR elements. Accessible through the EnvoFams portal, WasteFams provides a user-friendly interface featuring advanced search capabilities, sequence and structural similarity tools, and interactive visualization modules. As global initiatives increasingly leverage wastewater for public health and environmental insights, WasteFams can serve as a critical resource for discovering novel microbial functions, monitoring resistance mechanisms, and exploring the biotechnological potential of secondary metabolites within wastewater-engineered ecosystems.

bioinformatics↗

bioRxiv2026

NYX: Format-aware, learned compression across omics file types

Genomic data repositories continue to grow as sequencing technologies improve, with the NCBI SRA alone exceeding 47 PB. General-purpose compressors treat bioinformatics files as unstructured byte streams and fail to exploit the structured nature of omics data. We present NYX, a format-aware compression system for FASTA, FASTQ, VCF, WIG, H5AD, and BED files. NYX combines lightweight, reversible preprocessing and is build upon the OpenZL framework to take advantage of inherent data structure, delivering high compression ratios while preserving fast and lossless compression. Across representative datasets in the target formats, NYX achieves substantially higher speed than format-specific compressors while maintaining or improving compression ratio.

bioinformatics↗

bioRxiv2026

Fundamental limitations of genomic language models for realistic sequence generation

Large language models (LLMs) have shown remarkable success in natural language processing, prompting interest in their application to genomic sequence analysis. Genomic Language Models (gLMs) based on similar architectures offer a promising avenue for synthetic genome generation and characterization. However, their effectiveness for biological sequence modeling remains poorly characterized. We present a comprehensive evaluation of genomic language models that explicitly aim to generate entire synthetic genomes. We tested Evo 2 on diverse prokaryotic, eukaryotic and viral genomes, and megaDNA on bacteriophage genomes, and assessed performance across key biological features and organizational patterns. Our results reveal systematic failures in gLM-based genomic reconstruction. While the synthetic sequences captured local sequence statistics, they consistently failed to preserve long-range genomic organization, repeat and k-mer composition, transcription factor binding site architecture, and evolutionary constraints. Generated sequences exhibited violations of natural genomic patterns and models showed particular difficulty with repetitive elements. To assess the quality of genome generation, we trained a convolutional neural network that reliably distinguished synthetic from natural sequences, achieving AUROC values up to 0.97 in eukaryotes and 0.82 in prokaryotes, with classification accuracy increasing monotonically with genomic distance from the seed. These findings suggest fundamental limitations in current gLM architectures for capturing the long-range, hierarchical nature of genomic sequences. Our work highlights the need for specialized architectures that explicitly model evolutionary constraints rather than relying solely on statistical patterns, with important implications for computational biology applications requiring realistic sequence generation and for biosafety assessments that depend on the distinguishability of synthetic and natural genomic sequences.

synthetic biology↗

bioRxiv2026

Accelerating Inference in Genomic Foundation Models via Speculative Decoding

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. 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 models 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. 1.2x-1.4x), 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.

genomics↗

bioRxiv2025

Characterization of Z-DNA dynamics across the tree of life

Z-DNA/Z-RNA is an alternative left-handed nucleic acid conformation with established and emerging roles in gene regulation, immunity, and genome instability. However, its occurrence dynamics and lineage specificity across the tree of life have not yet been fully characterized. Utilizing the recently developed and improved Z-DNA searching tool, ZSeeker, we analyzed 281,139 complete organismal genomes, including multiple Telomere-to-Telomere genome assemblies, and generated genome-wide Z-nucleic acid maps, examined their topography, and compared them to dinucleotide-preserving controls. Cellular genomes featured pervasive Z-DNA enrichment relative to expectation, with enrichments of [~]1.5 and [~]1.7-fold in Bacteria and Archaea and [~]3-fold in Eukaryota. In contrast, Viruses exhibited large differences between lineages, with modest enrichment in several DNA viral groups and pronounced depletion across RNA clades, most notably Influenza A/B strains. We built a LASSO regression model trained on non-Influenza viruses (cross-validated R{superscript 2} {approx} 0.73), which identified GC content, genome type, and host type as the leading predictors for Z-nucleic acid density, yet it significantly over-predicted Z-RNA density in Influenza A/B. More than 99% of assemblies exceeded the +2 SD threshold, and a "typical Influenza" genome was predicted at 2.76 bp/kb compared to [~]0.016 bp/kb observed (a [~]170-fold overestimation based on chance alone). Together, these results reveal domain- and lineage-specific regimes: cellular genomes are enriched for Z-DNA consistent with regulatory roles, whereas influenza viruses appear to have undergone strong, lineage-specific depletion of Z-RNA-forming sequences, likely reflecting evolutionary pressure tied to host sensing pathways.

evolutionary biology↗

bioRxiv2025

Quadrupling the protein family space with global metagenomics

The known universe of protein families represents only a small fraction of natures molecular diversity. From 40.3 billion sequences across 40,446 metagenomes, 9,540 metatranscriptomes, and 539 million proteins from 167,415 reference genomes, we identified 608,258 previously uncharacterized protein families with [≥]100 members and 6.5 million families with [≥]25 members, none matching known Pfam domains or reference proteins. This effort doubles the known repertoire of large families and quadruples that of smaller families. Integration of AlphaFold2-based structural predictions with gene-neighborhood and taxonomic analyses enables the characterization of previously unannotated proteins, revealing candidates for both novel and known biological functions in understudied microbial lineages and biomes. This expanded repertoire provides insights into microbial adaptation and broadens the molecular toolkit available for biotechnology, highlighting the power of global metagenomics to uncover hidden protein diversity.

bioinformatics↗

bioRxiv2025

Quantum implementation of multi-pattern string matching for k-mer detection

MotivationThe exponential growth of publicly available genomic data has created unprecedented opportunities for sequence-based discovery. Locating specific k-mers is fundamental to diverse applications, including metagenomic classification, pathogen and cancer detection, and variant calling yet efficient identification of multiple k-mer patterns across large sequencing data and massive databases remains a significant computational challenge. MethodWe implement two quantum algorithms for DNA multi pattern string matching for k-mer detection based on Grovers amplitude amplification with quantum random access memory (QRAM). The first algorithm uses an enumerate-m oracle that sequentially checks a loaded text substring against all m patterns achieving O({surd}S) query complexity for S text positions but requiring O(m{middle dot}L) work per oracle call. The second algorithm employs nested Grover search with an outer loop over text positions and an inner loop over pattern space, reducing oracle complexity to O(L) while performing O({surd}S {middle dot} {surd}m) in total. ResultsWe present two quantum implementations of multi-pattern string matching tailored for k-mer detection. Leveraging quantum parallelism and Grover-inspired search primitives, our methods accelerate dictionary-based pattern matching, particularly in contexts involving large sequences, such as genomic data, and extensive pattern sets. ConclusionsWhile implementation challenges such as QRAM overhead remain, this study demonstrates both the promise and current limitations of quantum-enhanced string matching, establishing a foundational step toward quantum readiness in bioinformatics. Availability and implementationTo maximize accessibility and practical use, we provide our methodology at: https://github.com/Georgakopoulos-Soares-lab/quantum-multi-motif-finder

bioinformatics↗

bioRxiv2025

The cumulative impact of passenger mutations on cancer development

The traditional binary classification of somatic mutations in cancer as either drivers or passengers overlooks the potential cumulative impact of smaller-effect mutations. Here, we analyze 2,263 whole-genome-sequenced primary tumors across 31 cancer types to assess the functional contribution of passenger mutations in cancer development. We find that in the absence of canonical driver mutations, passenger mutations in cancer genes are significantly enriched, exhibit higher predicted pathogenicity, and are associated with aberrant expression, splicing disruption, altered transcription factor binding, and clinical outcomes that resemble those in the presence of driver mutations. The accumulation of passenger mutations in tumor suppressor genes correlates with significantly reduced expression and poorer prognosis, mirroring the functional outcomes of driver mutations. Notably, this is a previously uncharacterized mechanism of tumor suppressor inactivation, in which the accumulation of somatic mutations results in its progressive inactivation. Our results support a continuum model of mutational impact, where the collective influence of passenger mutations contributes to oncogenesis and clinical outcomes. This work advocates for integrative cancer models that incorporate all somatic mutations to more accurately reflect the complexity of tumor evolution.

cancer biology↗

bioRxiv2025

nf-core/proteinfamilies: A scalable pipeline for the generation of protein families

The growth of metagenomics-derived amino acid sequence data has transformed our understanding of protein function, microbial diversity and evolutionary relationships. However, the vast majority of these proteins remain functionally uncharacterized. Grouping the millions of such uncharacterised sequences with the few experimentally characterised ones allows the transfer of annotations, while the inspection of conserved residues with multiple sequence alignments can provide clues to function, even in the absence of existing functional information. To address the challenges associated with this data surge and the need to group sequences, we present a scalable, open-source, parametrizable Nextflow pipeline (nf-core/proteinfamilies) that generates protein nascent families or assigns new proteins to existing families. The computational benchmarks demonstrated that resource usage can scale approximately linearly with input size, while the biological benchmarks showed that the generated protein families closely resemble manually curated families found in widely used databases.

bioinformatics↗

bioRxiv2025

ZSeeker: An optimized algorithm for Z-DNA detection in genomic sequences

Z-DNA is an alternative left-handed helical form of DNA with a zigzag-shaped backbone that differs from the right-handed canonical B-DNA helix. Z-DNA has been implicated in various biological processes, including transcription, replication, and DNA repair, and can induce genetic instability. Repetitive sequences of alternating purines and pyrimidines have the potential to adopt Z-DNA structures. ZSeeker is a novel computational tool developed for the accurate detection of potential Z-DNA-forming sequences in genomes, addressing limitations of prior methods. By introducing a novel methodology informed and validated by experimental data, ZSeeker enables the refined detection of potential Z-DNA-forming sequences. Built both as a standalone Python package and as an accessible web interface, ZSeeker allows users to input genomic sequences, adjust detection parameters, and view potential Z-DNA sequence distributions and Z-scores via downloadable visualizations. Our Web Platform provides a no-code solution for Z-DNA identification, with a focus on accessibility, user-friendliness, speed and customizability. By providing efficient, high-throughput analysis and enhanced detection accuracy, ZSeeker has the potential to support significant advancements in understanding the roles of Z-DNA in normal cellular functions, genetic instability, and its implications in human diseases. AvailabilityZSeeker is released as a Python package under the GPL license as a multi-platform application and is available at: https://github.com/Georgakopoulos-Soares-lab/ZSeeker. A web-interface of ZSeeker is publicly available at https://zseeker.netlify.app/.

bioinformatics↗