Biology subjects

Bagordo, D.

Publications and source records attributed to Bagordo, D..

4 recordsLinked to original sources

bioRxiv2026

nanorepertoire: an end-to-end Nextflow pipeline for nanobody repertoire analysis

Camelid heavy-chain antibodies, and particularly their variable domains known as nanobodies or VHHs, combine full antigen-binding capacity with a compact and highly stable scaffold, which makes them attractive for both fundamental immunology and therapeutic development. High-throughput adaptive immune receptor repertoire sequencing (AIRR-seq) allows nanobody repertoires to be profiled at great depth, but the analyses applied to VHH data are typically assembled ad hoc from standalone scripts, which limits standardisation and reproducibility across laboratories. Here we present nanorepertoire, an end-to-end Nextflow DSL2 pipeline dedicated to camelid VHH repertoires. It takes paired-end AIRR-seq FASTQ files through quality control, adapter trimming, read merging, in-silico translation, CD-HIT clonotyping and deep-learning CDR3 annotation with nanoCDR-X (Bagordo et al., 2026), and returns an interactive HTML report describing clonal architecture, CDR3 length and amino-acid composition, intra-clonal homogeneity and repertoire diversity, together with the computational carbon footprint of the run. Applied to two publicly available SARS-CoV-2 RBD-selected llama libraries sampled before and after phage-display enrichment (4.8 million paired-end reads in total), the pipeline completed in 59 min on a 16-vCPU cloud instance and recovered 41,363 distinct CDR3 paratopes, reproducing the expected contraction of clonal diversity upon selection. nanorepertoire is open source under the MIT licence at https://github.com/lescailab/nanorepertoire, is archived on Zenodo, and runs unchanged on local, HPC and cloud infrastructures.

bioinformatics↗

bioRxiv2026

SIEVE: Sparse Interpretable Exome Variant Explainer

Whole-exome case-control studies contain rare and common variation, yet analytical methods usually partition the frequency spectrum, discard positional context, or depend on fixed annotations. We present SIEVE, a deep-learning framework for interpretable variant and gene prioritisation. It reads every observed exonic variant without a frequency filter, represents genomic position through self-attention, and calibrates attributions against a permuted-label null. Across coronary artery disease, early-onset myocardial infarction and Crohns disease, discrimination matches the liability-threshold expectation for each trait, while recovery of catalogued associations rises with annotation depth. Against burden testing, single-variant association and polygenic scoring, SIEVE recovers overlapping but largely distinct candidates.

bioinformatics↗

bioRxiv2025

Eduomics: a Nextflow pipeline to simulate -omics data for education

Moving past learning just algorithms and code is a key challenge of bioinformatics education: the ideal goal is for students to acquire higher-order knowledge such as the ability to solve biological problems with the appropriate tools, and more importantly learn to interpret the results in the broader context where bioinformatics is needed. To design such a teaching and learning experience, data simulations play a key role: however, there is a massive barrier to adoption. Different data types are produced by different tools, requiring educators to learn each of them and adapt their workflow to the necessary dependencies, requirements and input files. Additionally, most existing data simulation solutions are meant for benchmarking and methods development rather than education and cannot provide the context needed to teach students the critical interpretation skills they need to move beyond problem-based learning to what we call storyline-based learning. A significant effort must be placed also when many datasets with the same characteristics are needed, such as in tutoring or assessment in higher education. Here, we present eduomics: a Nextflow pipeline meant to automate the simulation workflow for both genomic and transcriptomic next-generation sequencing data, and to produce realistic clinical scenarios to provide students with clues and a biomedical story necessary for the interpretation of their results. Eduomics removes barriers to adoption, by requiring the user to just decide which chromosome datasets should be simulated on, and which type of data they would like to simulate. There is no need to learn specific tools and resolve their dependencies. The use of Gemini API provides an innovative approach to generate plausible clinical scenarios, consistent with the genes where either a pathological mutation or differential expression has been simulated. With eduomics, we offer an accessible and scalable solution to design comprehensive learning experiences and innovate bioinformatics education. Author SummaryModern bioinformatics education faces a dual challenge: students must learn how to analyse data and interpret results within the biological problems they aim to solve, while instructors who wish to design such a comprehensive learning experience should introduce appropriately simulated data in their teaching. However, the complexity of existing simulation tools is an often daunting barrier to overcome: current solutions require different software for different data types and provide no support for teaching interpretation skills. Most available simulators were designed for benchmarking and method development, not for education: therefore they lack the biological and clinical context needed to move beyond the traditional problem-based learning. To address these challenges, we developed eduomics, a Nextflow-based end-to-end pipeline that automates the simulation of genomic and transcriptomic data in a fully automated manner for educational use. Educators only need to choose the type of data to simulate, while the pipeline handles all underlying tools and dependencies. As a groundbreaking element, we integrated the Google Gemini API to generate patient-inspired clinical scenarios that gives students realistic clues and a biomedical narrative in which to interpret the results. This approach complements the traditional problem-based learning by offering a more complete, storyline-based learning experience. By combining automated data simulation and clinical storytelling, eduomics provides an accessible and scalable solution that support an engaging and immersive learning experience for both educators and students.

bioinformatics↗

bioRxiv2025

Adaptive disorder as the hallmark of nanobodies antigen-binding loops

Nanobodies are antigen-binding proteins of great interest as diagnostics and therapeutics. Accurate and fast characterization of their complementarity-determining regions (CDRs) is crucial to uncover the principles guiding their design. Yet, this task remains challenging, as random recombination and somatic mutations generate highly diverse CDR sequences that escape motif-based or structure-prediction approaches currently used to identify them. To overcome this hurdle, we employed two independent strategies that converged on the same conclusion. At the sequence level, we developed a deep learning model to identify nanobody CDRs directly from the primary sequence. At the structural level, we applied an energy decomposition method, revealing CDRs as residues highly uncoupled to the rest of the fold. Explainability analyses showed the network captured intrinsic CDR properties, which notably aligned with these energy values. CDRs emerge as fuzzy regions capable of adopting diverse conformational ensembles, from which a preferred state is selected upon antigen binding. This finding supports a model where chaos in both sequence and structure appears adaptive and disorder emerges as the hallmark of nanobody CDRs. This work aims to advance the definition of rules for the design of antigen binding regions, paving the way for the next-generation immune diagnostics and therapeutics.

immunology↗