Biology subjects

Alexiou, P.

Publications and source records attributed to Alexiou, P..

16 recordsLinked to original sources

bioRxiv2026

Argonaute 1 associates with nucleolar snoRNP machinery and regulates pre-rRNA levels and cell growth

Argonaute proteins are best known as cytoplasmic effectors of small-RNA-mediated gene silencing, although their compartment-specific functions remain incompletely understood. Here, we identify a previously unrecognized nucleolar role for Drosophila Argonaute 1 (AGO1). We observed stress-induced redistribution of AGO1 into cytoplasmic and nuclear ribonucleoprotein assemblies, whereas a distinct AGO1 pool consistently localized to Fibrillarin-positive nucleoli. Co-immunoprecipitation further showed that AGO1 associates with Fibrillarin-containing complexes in vivo. Re-analysis of AGO1-associated small-RNA sequencing data revealed snoRNAs, particularly C/D box snoRNAs, among AGO1-bound RNA species and increased snoRNA representation following AGO1 overexpression. Structural modelling demonstrated that simultaneous molecular interaction of AGO1 and the Fibrillarin-containing snoRNP machinery with the same snoRNA is stereochemically feasible, providing a structural framework for their potential association within a nucleolar snoRNP assembly. Functionally, AGO1 depletion reduced pre-rRNA levels and decreased both nucleolar and cell size in the follicular epithelium. Depletion of Dicer-1 did not reproduce these growth defects, arguing against a general consequence of impaired miRNA biogenesis. Together, these findings identify AGO1 as a nucleolar regulator that associates with snoRNP machinery and links small-RNA biology to pre-rRNA homeostasis and cell growth.

cell biology↗

bioRxiv2026

pre-miRBench: a multispecies benchmark and reference model for animal precursor microRNA prediction

Motivation: Precursor microRNA (pre-miRNA) prediction is used to prioritize genomic hairpins for miRNA annotation. Published predictors are trained and evaluated on different species collections, negative sets and data partitions, making reported gains difficult to compare. Results: We introduce pre-miRBench, a benchmark of 7,056 200-nt windows centered on curated pre-miRNA loci and 70,560 genomic negatives with predicted hairpin structures from 71 animal species, with a fixed 1:10 positive-to-negative ratio in every partition. Four test sets cross whether the species and miRNA family are represented in training: known species/known family, known species/held-out family, held-out species/known family, and held-out species/held-out family. We retrained six published pre-miRNA predictors (DeepMir, deepMiRGene, dnnPreMiR, mirDNN, miRe2e and MuStARD) under the same benchmark splits and evaluated them alongside the pre-miRBench model, an ensemble of three neural networks developed with Agentomics, an autonomous LLM-agent system for biomedical machine learning. The pre-miRBench model achieved the highest mean average precision (AP) across the four test sets (0.9809; dnnPreMiR, 0.9719) and led three tests. The two held-out species were Gallus gallus and Drosophila melanogaster, and the joint species-and-family holdout contained 69 positives and 690 negatives. dnnPreMiR had the highest AP on the joint species-and-family holdout (0.9677 versus 0.9659). The AP difference (pre-miRBench minus dnnPreMiR) was -0.0018, with a paired bootstrap 95% confidence interval of -0.0445 to 0.0341. pre-miRBench establishes a common evaluation for species and miRNA-family holdouts and provides a reproducible reference model for future comparisons. Availability and implementation: Reproducible source code is available at https://github.com/dimostzim/pre-miRBench. Datasets, trained model weights, record-level predictions, evaluation metrics and species-panel metadata for pre-miRBench and the six retrained predictors are available at https://zenodo.org/records/22813044.

bioinformatics↗

bioRxiv2026

ParaDISM: Precise mapping of short reads to genes with highly homologous regions

BackgroundGenes with highly similar genomic copies (paralogs, tandem duplications and pseudogenes) pose a major challenge for Short-Read High Throughput Sequencing (srHTS). High sequence similarity makes it difficult to unambiguously identify the sequences of origin of short reads. This results in misalignment artifacts which can propagate through bioinformatic pipelines and increase error rates in variant calling. ResultsWe present ParaDISM, a pipeline that refines standard alignments to improve read placement and reduce misalignment-driven false variant calls in highly homologous sequences. ParaDISM assigns a read/read pair to a sequence only when supported by unambiguous sequence-specific evidence by using a multiple sequence alignment of reference sequences to identify disambiguating positions. An optional iterative refinement procedure calls variants from confidently assigned reads, updates the reference sequences, and processes remaining non-assigned reads. We evaluated the performance of ParaDISM both in terms of read alignment and the resulting short variant calls using extensive computational simulation experiments and the Genome in a Bottle HG002 benchmark. We applied ParaDISM to reanalyze two case studies: five public tumour exomes at the GNAQ/GNAQP1 locus, and 18 short-read sequencing datasets of patients diagnosed with Autosomal Dominant Polycystic Kidney Disease (16 exomes and 2 panel sequencing datasets). Compared to the standard aligners (bowtie2, bwa-mem and minimap2), ParaDISM reduced the number of misalignment artifacts and false variant calls, resulting in an increased specificity and precision of the results. ConclusionsParaDISM improves the precision of read placement and single-nucleotide variant calling in highly homologous reference sequences. By reducing the number of false variant calls caused by misalignment artifacts, ParaDISM provides a stronger level of evidence for the called variants compared to currently available approaches. The pipeline is open source and available under the MIT license at github.com/BioGeMT/ParaDISM.

bioinformatics↗

bioRxiv2026

Discriminative learning of substitution matrices and gap penalties for pairwise alignment of biological sequences

Pairwise alignment scores are used to classify pairs of sequences in many areas of bioinformatics, including homology search, predicting interactions, or read mapping. The relative scores of different pairs strongly depend on the choice of a substitution matrix and gap penalties, but the existing approaches for the estimation of these parameters do not directly optimize them for the task of classification. In this work, we present DiscrimAlign, a statistical model for discriminative learning of substitution matrices and gap penalties from a dataset of positive and negative pairs of unaligned biological sequences. The model links the alignment score of a sequence pair with the associated binary label through a logistic function and learns the parameters by likelihood maximization. We analyze theoretical properties of the model, derive and implement a learning procedure, study its performance in simulated experiments, and apply it to predict microRNA-target interactions. We show that sequence alignment with discriminative substitution matrices and gap penalties predicts the interactions comparably to state-of-the-art neural network classifiers while being more interpretable. An implementation of the model and reproducibility workflows are available at https://github.com/BioGeMT/DiscrimAlign.

bioinformatics↗

bioRxiv2026

miRBind2 enables sequence-only prediction of miRNA binding and transcript repression

MotivationMicroRNAs (miRNAs) regulate gene expression by guiding Argonaute proteins to partially complementary sites on target RNAs. While classical prediction methods rely on engineered features such as seed match categories, evolutionary conservation, and site context, recent advances in deep learning offer the potential to learn targeting rules directly from sequence. We developed a sequence-based deep learning model that improves miRNA target site prediction, and further validated the learned target site representations by extending the model to gene-level functional repression prediction. ResultsWe introduce miRBind2, a deep learning method for miRNA target site prediction that incorporates a novel pairwise nucleotide representation capturing all possible miRNA-target nucleotide interactions, with a CNN-based architecture. miRBind2 outperforms previous SotA models across four independent datasets from the debiased miRBench benchmark, while using 92% fewer parameters. We show that the convolutional features and weights learned by miRBind2 can be transferred to transcript-level prediction by extending the miRBind2 architecture and fine-tuning it on miRNA perturbation experiments. This miRBind2-3UTR model predicts gene repression from sequence alone. On a dataset of 50,549 miRNA-gene pairs, miRBind2-3UTR significantly outperforms TargetScan. These results show that deep models pretrained on target site data can capture regulatory signals and predict functional repression without requiring conventional engineered biological features. AvailabilityModels and source code are freely available via GitHub (https://github.com/BioGeMT/miRBind_2.0). A publicly available web-tool for novel predictions and visualization is available at : (https://huggingface.co/spaces/dimostzim/BioGeMT-miRBind2) Contactpanagiotis.alexiou@um.edu.mt

bioinformatics↗

bioRxiv2026

Agentomics: An Agentic System that Autonomously Develops Novel State-of-the-art Solutions for Biomedical Machine Learning Tasks

MotivationExtracting knowledge from biomedical data is crucial for advancing our understanding of biological systems and developing novel therapeutics. The quantity, quality, and resolution of biomedical data constantly evolves, requiring the automation of biomedical machine learning (ML). Existing Automated ML tools lack flexibility, while Large Language Models (LLMs) struggle to consistently deliver reproducible machine learning codebases, and existing LLM Agent-powered solutions lag behind human-engineered ML models. ResultsHere, we introduce Agentomics, an autonomous LLM-powered agentic system for end-to-end ML experimentation. Given a biomedical dataset, Agentomics implements various ML modeling strategies, and produces a ready-to-use ML model. Agentomics introduces strict validation checkpoints for standard ML development steps, allowing gradual development on top of working code with defined interfaces and validated artifacts. Further, it offers native support for biomedical foundation models that can be leveraged during experimentation. The generic nature of Agentomics allows the user to create ML solutions for a large variety of datasets and use various LLMs. We evaluate Agentomics across 20 datasets from the domains of Protein Engineering, Drug Discovery, and Regulatory Genomics. When benchmarked against other agentic systems, Agentomics outperformed them in all tested domains. When benchmarked against human expert solutions, Agentomics generated novel state-of-the-art models for 11/20 established benchmark datasets. Availability and ImplementationAgentomics is implemented in Python. Source code and documentation are freely available at: https://github.com/BioGeMT/Agentomics-ML. Contactpanagiotis.alexiou@um.edu.mt

bioinformatics↗

bioRxiv2024

miRBench: A Comprehensive microRNA Binding Site Prediction Training and Benchmarking Dataset

MotivationMicroRNAs (miRNAs) are crucial regulators of gene expression, but the precise mechanisms governing their binding to target sites remain unclear. A major contributing factor to this is the lack of unbiased experimental datasets for training accurate prediction models. While recent experimental advances have provided numerous miRNA-target interactions, these are solely positive interactions. Generating negative examples in silico is challenging and prone to introducing biases, such as the miRNA frequency class bias identified in this work. Biases within datasets can compromise model generalization, leading models to learn dataset-specific artifacts rather than true biological patterns. ResultsWe introduce a novel methodology for negative sample generation that effectively mitigates the miRNA frequency class bias. Using this methodology, we curate several new, extensive datasets and benchmark several state-of-the-art methods on them. We find that a simple convolutional neural network model, retrained on some of these datasets, is able to outperform state-of-the-art methods. This highlights the potential for leveraging unbiased datasets to achieve improved performance in miRNA binding site prediction. To facilitate further research and lower the barrier to entry for machine learning researchers, we provide an easily accessible Python package, miRBench, for dataset retrieval, sequence encoding, and the execution of state-of-the-art models. AvailabilityThe miRBench Python Package is accessible at https://github.com/katarinagresova/miRBench/releases/tag/v1.0.0

bioinformatics↗

bioRxiv2023

Deep learning and direct sequencing of labeled RNA captures transcriptome dynamics

Quantification of the dynamics of RNA metabolism is essential for understanding gene regulation in health and disease. Existing methods rely on metabolic labeling of nascent RNAs and physical separation or inference of labeling through PCR-generated mutations, followed by short-read sequencing. However, these methods are limited in their ability to identify transient decay intermediates or co-analyze RNA decay with cis-regulatory elements of RNA stability such as poly(A) tail length and modification status, at single molecule resolution. Here we use 5-ethynyl uridine (5EU) to label nascent RNA followed by direct RNA sequencing with nanopores. We developed RNAkinet, a deep convolutional and recurrent neural network that processes the electrical signal produced by nanopore sequencing to identify 5EU-labeled nascent RNA molecules. RNAkinet demonstrates generalizability to distinct cell types and organisms and reproducibly quantifies RNA kinetic parameters allowing the combined interrogation of RNA metabolism and cis-acting RNA regulatory elements.

bioinformatics↗

bioRxiv2023

Beyond microRNAs: Analysis of chimeric reads characterises the diverse targetome of AGO2-mediated regulation.

Argonaute proteins are instrumental in regulating RNA stability and translation. AGO2, the major mammalian Argonaute protein, is known to primarily associate with microRNAs, a family of small RNA driver sequences, and identifies its targets primarily via a seed mediated partial complementarity process Despite numerous studies, a definitive experimental dataset of AGO2 driver-target interactions remains elusive. Our study employs two experimental methods - AGO2 CLASH and AGO2 eCLIP, to generate thousands of AGO2 target sites verified by chimeric reads. These chimeric reads contain both the AGO2 loaded small RNA driver and the target sequence, providing a robust resource for modeling AGO2 binding preferences. Our novel analysis pipeline reveals thousands of AGO2 target sites driven by microRNAs and a significant number of AGO2 drivers derived from fragments of other small RNAs such as tRNAs, YRNAs, snoRNAs, rRNAs, and more. We utilize convolutional neural networks to train machine learning models that accurately predict the binding potential for each driver class and experimentally validate several interactions. In conclusion, our comprehensive analysis of the AGO2 targetome broadens our understanding of its driver repertoire and potential function in development and disease. Moreover, we offer practical bioinformatic tools for future experiments and the prediction of AGO2 targets. All data and code from this study are freely available at https://github.com/ML-Bioinfo-CEITEC/HybriDetector/ Contactpanagiotis.alexiou@um.edu.mt, stepanka.vanacova@ceitec.muni.cz

molecular biology↗

bioRxiv2023

SAP domain facilitates efficient loading of Ku onto DNA ends

Recognition and processing of DNA ends play a central role in maintaining genome integrity. The evolutionarily conserved DNA repair complex Ku serves as the primary sensor of free DNA ends in eukaryotic cells. Its rapid association with DNA ends is crucial for several cellular processes, including non-homologous end joining (NHEJ) DNA repair and telomere protection. In this study, we conducted a transient kinetic analysis to investigate the impact of the SAP domain on individual phases of the Ku-DNA interaction. Specifically, we examined the initial binding, the subsequent docking of Ku onto DNA, and the sliding of Ku along DNA. Our findings revealed that the C-terminal domain of Ku70, known as SAP ((SAF-A/B, Acinus and PIAS), facilitates the initial phases of Ku-DNA interaction, but does not affect the sliding process. This suggests that SAP may either establish the first interactions with DNA, or stabilize these initial interactions during loading. To assess the biological role of SAP, we generated Arabidopsis plants expressing Ku lacking the SAP domain ({Delta}SAP). Intriguingly, despite the decreased efficiency of the {Delta}SAP Ku complex in loading onto DNA, the mutant plants exhibited full proficiency in classical NHEJ and telomere maintenance. This indicates that the speed of Ku loading onto telomeres or DNA double-strand breaks (DSBs) is not the decisive factor in stabilizing these DNA structures.

molecular biology↗

bioRxiv2022

The MOV10 RNA helicase is a dosage-dependent host restriction factor for LINE1 retrotransposition in mice

Transposable elements constitute nearly half of the mammalian genome and play important roles in genome evolution. While a multitude of both transcriptional and post-transcriptional mechanisms exist to silence transposable elements, control of transposition in vivo remains poorly understood. MOV10, an RNA helicase, is a potent inhibitor of mobilization of retrotransposons and retroviruses in cell culture assays. Here we report that MOV10 restricts LINE1 retrotransposition in mice. Although MOV10 is broadly expressed, its loss causes only incomplete penetrance of embryonic lethality, and the surviving MOV10-deficient mice are healthy and fertile. Biochemically, MOV10 forms a complex with UPF1, a key component of the nonsense-mediated mRNA decay pathway, and primarily binds to the 3UTR of somatically expressed transcripts in testis. Consequently, loss of MOV10 results in an altered transcriptome and a modest upregulation of two LINE1 families in testis. Analyses using a LINE1 reporter transgene reveal that loss of MOV10 leads to increased LINE1 retrotransposition in somatic and reproductive tissues from both embryos and adult mice. Moreover, the degree of LINE1 retrotransposition inhibition is dependent on the Mov10 gene dosage. Furthermore, MOV10 deficiency reduces reproductive fitness over successive generations. Our findings demonstrate that MOV10 attenuates LINE1 retrotransposition in a dosage-dependent manner in mice. Author summaryTransposable elements (TEs), including L1 and SINEs, are abundant in the genome and play important roles in evolution, development, and diseases. While TEs propagate in individuals and across generations, the host organism needs to suppress them, resulting in an ongoing arms race between TEs and the host genome. L1, a retrotransposon, accounts for about 17% of the mammalian genome. L1 encodes two proteins, which bind to the L1 transcript to form L1 ribonucleoprotein particles. L1 proliferates in the genome via retrotransposition. A multitude of transcriptional and post-transcriptional mechanisms exist to suppress TEs, however, retrotransposition of TEs remains poorly understood. L1 ribonucleoprotein particles are associated with a large number of host proteins, one of which is the MOV10 RNA helicase. MOV10 exhibits anti-viral activities against retroviruses such as HIV-1. In cultured cells, MOV10 is a potent inhibitor of retrotransposition of L1, SINEs, and IAP. Although MOV10 is expressed in a broad range of tissues, loss of MOV10 causes only incomplete penetrance of embryonic lethality. The viable MOV10-deficient mice are grossly normal and fertile. Importantly, analyses using a L1 transgene reporter reveal that MOV10 inhibits L1 retrotransposition in both somatic tissues and reproductive tissues in a gene dosage-dependent manner. Therefore, MOV10 functions as a host restriction factor for L1 and possibly other transposable elements in vivo.

developmental biology↗

bioRxiv2022

Genomic Benchmarks: A Collection of Datasets for Genomic Sequence Classification

In this paper, we propose a collection of curated and easily accessible sequence classification datasets in the field of genomics. The proposed collection is based on a combination of novel datasets constructed from the mining of publicly available databases and existing datasets obtained from published articles. The main aim of this effort is to create a repository for shared datasets that will make machine learning for genomics more comparable and reproducible while reducing the over-head of researchers that want to enter the field. The collection currently contains eight datasets that focus on regulatory elements (promoters, enhancers, open chromatin region) from three model organisms: human, mouse, and roundworm. A simple convolution neural network is also included in a repository and can be used as a baseline model. Benchmarks and the baseline model are distributed as the Python package genomic-benchmarks, and the code is available at https://github.com/ML-Bioinfo-CEITEC/genomic_benchmarks.

bioinformatics↗

bioRxiv2022

Fine-Tuning Transformers For Genomic Tasks

Transformers are a type of neural network architecture that has been successfully used to achieve state-of-the-art performance in numerous natural language processing tasks. However, what about DNA, the language life written in the four-letter alphabet? In this paper, we review the current state of Transformers usage in genomics and molecular biology in general, introduce a collection of benchmark datasets for the classification of genomic sequences, and compare the performance of several model architectures on those benchmarks, including a BERT-like model for DNA sequences DNABERT as implemented in HuggingFace (armheb/DNA_bert_6 model). In particular, we explore the effect of pre-training on a large DNA corpus vs training from scratch (with randomized weights). The results presented here can be used for identification of functional elements in human and other genomes.

bioinformatics↗

bioRxiv2021

ENNGene: an Easy Neural Network model building tool for Genomics

BackgroundThe recent big data revolution in Genomics, coupled with the emergence of Deep Learning as a set of powerful machine learning methods, has shifted the standard practices of machine learning for Genomics. Even though Deep Learning methods such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) are becoming widespread in Genomics, developing and training such models is outside the ability of most researchers in the field. ResultsHere we present ENNGene - Easy Neural Network model building tool for Genomics. This tool simplifies training of custom CNN or hybrid CNN-RNN models on genomic data via an easy-to-use Graphical User Interface. ENNGene allows multiple input branches, including sequence, evolutionary conservation, and secondary structure, and performs all the necessary preprocessing steps, allowing simple input such as genomic coordinates. The network architecture is selected and fully customized by the user, from the number and types of the layers to each layer's precise set-up. ENNGene then deals with all steps of training and evaluation of the model, exporting valuable metrics such as multi-class ROC and precision-recall curve plots or TensorBoard log files. To facilitate interpretation of the predicted results, we deploy Integrated Gradients, providing the user with a graphical representation of an attribution level of each input position. To showcase the usage of ENNGene, we train multiple models on the RBP24 dataset, quickly reaching the state of the art while improving the performance on more than half of the proteins by including the evolutionary conservation score and tuning the network per protein. ConclusionsAs the role of DL in big data analysis in the near future is indisputable, it is important to make it available for a broader range of researchers. We believe that an easy-to-use tool such as ENNGene can allow Genomics researchers without a background in Computational Sciences to harness the power of DL to gain better insights into and extract important information from the large amounts of data available in the field. Contactpanagiotis.alexiou@ceitec.muni.cz

bioinformatics↗

bioRxiv2020

PENGUINN: Precise Exploration of Nuclear G-quadruplexes Using Interpretable Neural Networks

G-quadruplexes (G4s) are a class of stable structural nucleic acid motifs that are known to play a role in a wide spectrum of genomic functions, such as DNA replication and transcription. The classical understanding of G4 structure points to four variable length guanine strands joined by variable length stretches of other nucleotides. Experiments using G4 immunoprecipitation and sequencing experiments have produced a high number of highly probable G4 forming genomic sequences. The expense and technical difficulty of experimental techniques highlights the need for computational approaches of G4 identification. Here, we present PENGUINN, a machine learning method based on Convolutional Neural Networks, that learns the characteristics of G4 sequences and accurately predicts G4s outperforming the state-of-the-art. We provide both a standalone implementation of the trained model, and a web application that can be used to evaluate sequences for their G4 potential.

bioinformatics↗

bioRxiv2019

MuStARD: Deep Learning for intra- and inter-species scanning of functional genomic patterns

Regions of the genome that produce different classes of functional elements also exhibit different patterns in their sequence, secondary structure, and evolutionary conservation. Deep Learning is a family of Machine Learning algorithms recently applied to a variety of pattern recognition problems. Here we present MuStARD (gitlab.com/RBP_Bioinformatics/mustard) a Deep Learning framework that can learn and combine sequence, structure, and conservation patterns in sets of functional regions, and accurately identify additional members of the given set over wide genomic areas. MuStARD is designed with general use in mind, and has sophisticated iterative fully-automated background selection capability. We demonstrate that MuStARD can be trained without changes on different classes of human small RNA loci (pre-microRNAs and snoRNAs) and accurately build prediction models for both, outperforming state of the art methods specifically designed for each specific class. Furthermore, we demonstrate the ability of MuStARD for inter-species identification of functional elements by predicting mouse small RNAs using human trained models. MuStARD is easy to deploy and extend to a variety of genomic classification questions.

bioinformatics↗