Biology subjects

Romano, S.

Publications and source records attributed to Romano, S..

11 recordsLinked to original sources

bioRxiv2026

Interpreting Graph Neural Networks for Antibacterial Drug Repurposing against Acinetobacter baumannii

Antimicrobial resistance in Acinetobacter baumannii, whose carbapenem-resistant strains are classified by the World Health Organization as a critical priority, has motivated machine-learning screens for new antibacterials, but translating predictions into chemical hypotheses requires evaluating the molecular features on which a model relies. A graph neural network ensemble achieves competitive predictive performance against a descriptor-augmented reference model on the same antibacterial screen while using only atom features and connectivity, allowing interpretation to focus on the molecular graph without a separate input of handcrafted descriptors. The atom-level explanations identify features supporting the model's activity predictions under perturbation and restoration, while comparisons with equally sized perturbations within each molecule distinguish piperazine as the most consistently supported recurring contributor to predicted activity. Prospective screening of the Drug Repurposing Hub followed by developability and safety assessment yields a ranked shortlist for experimental follow-up and leads to testing of CE3F4, which reduced A. baumannii growth by 57% at 0.25 mM in a preliminary assay with a non-monotonic concentration--response profile.

bioinformatics↗

bioRxiv2026

LLMsFold: Integrating Large Language Models and Biophysical Simulations for De Novo Drug Design

The discovery of novel small molecules is challenging because of the vastness of chemical space and the complexity of protein-ligand interactions, leading to low success rates and time-consuming workflows. Here, we present LLMsFold, a computational framework that combines Large Language Models (LLMs) and biophysical foundation tools to design and validate new small molecules targeting pathogenic proteins. The pipeline starts by identifying viable binding pockets on a target protein through geometry-based pocket detection. A 70-billion-parameter transformer model from the LlaMA family then generates candidate molecules as SMILES strings under prompt constraints that enforce drug-likeness. Each molecule is evaluated by Boltz-2, a diffusion-based model for protein-ligand co-folding that predicts bound 3D structure and binding affinity. Promising candidates are iteratively optimized through a reinforcement learning loop that prioritizes high predicted affinity and synthetic accessibility. We demonstrate the approach on two challenging targets: ACVR1 (Activin A Receptor Type 1), implicated in fibrodysplasia ossificans progressiva (FOP), and CD19, a surface antigen expressed on most B-cell lymphoma and leukemia cells. Top candidates show strong in silico binding predictions and favorable drug-like profiles. All code and models are made available to support reproducibility and further development.

bioinformatics↗

bioRxiv2023

Machine learning-based meta-analysis reveals gut microbiome alterations associated with Parkinson's disease

There is strong interest in exploring the potential of the gut microbiome for Parkinsons disease (PD) diagnosis and treatment. However, a consensus on the microbiome features associated with PD and a multi-study assessment of their diagnostic value is lacking. Here, we present a machine learning meta-analysis of PD microbiome studies of unprecedented scale (including 4,490 samples). Within most studies, microbiome-based machine learning models could accurately classify PD patients. However, models were study-specific and did not generalise well across other studies. By training models on multiple datasets, we could improve their general applicability and disease specificity as assessed against microbiomes from other neurodegenerative diseases. Meta-analysis of shotgun metagenomes moreover delineated PD-associated microbial pathways potentially contributing to the deterioration of gut health and favouring the translocation of pathogenic molecules along the gut-brain axis. Strikingly, diverse microbial pathways for the biotransformation of solvents and pesticides were enriched in PD. These results align with the epidemiological evidence that exposure to these molecules increases PD risk and raise the question of whether gut microbial metabolism modulates their toxicity. Taken together, we offer the most comprehensive overview to date about the PD gut microbiome and provide future reference for its diagnostic and functional potential.

microbiology↗

bioRxiv2022

Circulating U13 small nucleolar RNA as a candidate biomarker for Huntington's disease

Background and ObjectivesFluid biomarkers are a recent field of interest in Huntington disease (HD). We focused on small circulating RNAs from plasma of subjects with prodromal (pre-HD) and overt disease by a two-stage approach: an unbiased investigation by an array method and a validation study to quantify a significant small nucleolar RNA. MethodsThrough Affymetrix Gene-Chip-miRNA-Array we performed an exploratory study on 9 HD patients, 8 healthy subjects (HS) and 5 psychiatric patients (PP; who share drugs with HD patients, to control for iatrogenic effects). Through real time PCR we validated the results in an independent population of 24 HD patients, 15 pre-HD, 24 PP, 28 Alzheimers disease (AD) patients (added to control the disease-specificity of our finding) and 23 HS. A bioinformatic analysis was also performed to interpret our finding. ResultsThe microarray results showed a significant signal for U13 small nucleolar RNA (SNORD13) that was increased in plasma of HD patients compared to controls (fold change, 1.54, p =0.003 HD vs. HS, and fold change 1.44 p = 0.0026 HD vs. PP). In the validation population the significant increase in HD patients was evident compared to both pre-HD and the three control groups (p<0.00001). The plasma levels of SNORD13 correlated with the status of mutant huntingtin carrier and the disease duration (respectively R=0.69; p<0.000001; R=0.49; p=0.015). Through receiver operating characteristic (ROC) curve analysis, we showed high accuracy of plasmatic SNORD13 in discriminating HD patients from pre-HD and control groups (AUC=0.963), outperforming values reported in another study for intrathecal or plasmatic mutant huntingtin and neurofilament light chain as biomarkers of overt HD. The bioinformatic analysis on SNORD13 interactome and pathway analysis showed enrichments for factors involved in nuclear functions beyond the ribosome biogenesis. DiscussionWe report the unprecedented finding of a potential role of small nucleolar RNAs in HD. Circulating SNORD13 seems a good biomarker for clinical purposes. It seems to be specific for HD and to peripherally report a plausible tipping point in the pathogenic cascade at neuronal level, possibly paving the way for new therapeutic targets.

neuroscience↗

bioRxiv2021

Enrichment analysis of GWAS data in autoimmunity delineates the multiple sclerosis-Epstein Barr virus association

We exploited genetic information to assess non-genetic influences in autoimmunity. We isolated gene modules whose products physically interact with environmental exposures related to autoimmunity, and analyzed their nominal statistical evidence of association with autoimmune and non-autoimmune diseases in genome-wide association studies (GWAS) data. Epstein Barr virus (EBV) and other Herpesviruses interactomes emerged as specifically associated with multiple sclerosis (MS), possibly under common regulatory mechanisms. Analyses of MS blood and brain transcriptomes, cytofluorimetric studies of endogenous EBV-infected lymphoblastoid lines, and lesion immunohistochemistry, confirmed a dysregulation of MS-associated EBV interactors, suggesting their contribution to CD40 signaling alterations in MS. These interactors resulted enriched in modules from inherited axonopathies-causing genes, supporting a link between EBV and neurodegeneration in MS, in accord with the observed transcriptomic dysregulations in MS brains. They were also enriched with top-ranked pharmaceutical targets prioritized on a genetic basis. This study delineates a disease-specific influence of herpesviruses on MS biology.

neuroscience↗

bioRxiv2021

GWAS-associated variants, non-genetic factors and transient transcriptome in Multiple Sclerosis etiopathogenesis: a colocalization analysis

A clinically actionable understanding of multiple sclerosis (MS) etiology goes through GWAS interpretation, prompting research on new gene regulatory models. We previously suggested a stochastic etiologic model where small-scale random perturbations could reach a threshold for MS development. The recently described mapping of the transient transcriptome (TT), including intergenic and intronic RNAs, seems appropriate to verify this model through a rigorous colocalization analysis. We show that genomic regions coding for the TT were significantly enriched for MS-associated GWAS variants and DNA binding sites for molecular transducers mediating putative, non-genetic, etiopathogenetic factors for MS (e.g., vitamin D deficiency, Epstein Barr virus latent infection, B cell dysfunction). These results suggest a model whereby TT-coding regions are hotspots of convergence between genetic ad non-genetic factors of risk/protection for MS, and plausibly for other complex disorders. Our colocalization analysis also provides a freely available data resource (www.mscoloc.com) for future research on MS transcriptional regulation.

neuroscience↗

bioRxiv2020

The hidden pangenome: comparative genomics reveals pervasive diversity in symbiotic and free-living sulfur-oxidizing bacteria

Sulfur-oxidizing Thioglobaceae, often referred to as SUP05 and Arctic96BD clades, are widespread and common to hydrothermal vents and oxygen minimum zones. They impact global biogeochemical cycles and exhibit a variety of host-associated and free-living lifestyles. The evolutionary driving forces that led to the versatility, adoption of multiple lifestyles and global success of this family are largely unknown. Here, we perform an in-depth comparative genomic analysis using all available and newly generated Thioglobaceae genomes. Gene content variation was common, throughout taxonomic ranks and lifestyles. We uncovered a pool of variable genes within most Thioglobaceae populations in single environmental samples and we referred to this as the hidden pangenome. The hidden pangenome is often overlooked in comparative genomic studies and our results indicate a much higher intra-specific diversity within environmental bacterial populations than previously thought. Our results show that core-community functions are different from species core genomes suggesting that core functions across populations are divided among the intra-specific members within a population. Defense mechanisms against foreign DNA and phages were enriched in symbiotic lineages, indicating an increased exchange of genetic material in symbioses. Our study suggests that genomic plasticity and frequent exchange of genetic material drives the global success of this family by increasing its evolvability in a heterogeneous environment.

microbiology↗

bioRxiv2020

Loss of dmrt1 restores female fates in the absence of cyp19a1a but not rbpms2

Sex determination and differentiation is a complex process regulated by multiple factors, including factors from the germline or surrounding somatic tissue. In zebrafish, sex-determination begins with establishment of a bipotential gonad that undergoes sex-specific differentiation and maintenance to form the functional adult gonad. However, the relationships among these factors are not fully understood. Here we identify potential Rbpms2 targets and apply genetic epistasis experiments to decipher the genetic hierarchy of regulators of sex-specific differentiation. We provide evidence that the critical female factor, rbpms2 is epistatic to the male factor dmrt1 in terms of adult sex. Moreover, Rbpms2s role in promoting female fates extends beyond repression of Dmrt1, as Rbpms2 is essential for female differentiation even in the absence of Dmrt1. In contrast, female fates can be restored in mutants lacking cyp19a1a in the absence of dmrt1. Taken together this work indicates that Cyp19a1a-mediated suppression of Dmrt1 is key to establish a bipotential gonad and initiate female fate acquisition, possibly by promoting rbpms2. Then, after female fate specification, Cyp19a1a regulates subsequent oocyte maturation and sustains female fates independent of Dmrt1 repression. Author SummaryWe show that cyp19a1a-mediated suppression of dmrt1 establishes a bipotential gonad and female fate acquisition, possibly through rbpms2 which is required for female fates, even in the absence of Dmrt1.

developmental biology↗

bioRxiv2019

Diversity matters: Deep-sea mussels harbor multiple symbiont strains

Genetic diversity of closely-related free-living microbes is widespread and underpins ecosystem functioning, but most evolutionary theories predict that it destabilizes intimate mutualisms. Indeed, symbiont strain diversity has long assumed to be restricted in intracellular bacteria associated with animals. Here, we sequenced the metagenomes and metatranscriptomes of 18 Bathymodiolus mussel individuals from four species, covering their known distribution range at deep-sea hydrothermal vents in the Atlantic. We show that as many as 16 strains of intracellular, sulfur-oxidizing symbionts coexist in individual Bathymodiolus mussels. Co-occurring symbiont strains differed extensively in key metabolic functions, such as the use of energy and nutrient sources, electron acceptors and viral defense mechanisms. Most strain-specific genes were expressed, highlighting their adaptive potential. We show that fine-scale diversity is pervasive in Bathymodiolus symbionts, and hypothesize that it may be widespread in low-cost symbioses where the environment, not the host, feeds the symbionts.

microbiology↗

bioRxiv2018

The genome of the jellyfish Clytia hemisphaerica and the evolution of the cnidarian life-cycle

Jellyfish (medusae) are a distinctive life-cycle stage of medusozoan cnidarians. They are major marine predators, with integrated neurosensory, muscular and organ systems. The genetic foundations of this complex form are largely unknown. We report the draft genome of the hydrozoan jellyfish Clytia hemisphaerica and use multiple transcriptomes to determine gene use across life-cycle stages. Medusa, planula larva and polyp are each characterised by distinct transcriptome signatures reflecting abrupt life cycle transitions, and all deploy a mixture of phylogenetically old and new genes. Medusa specific transcription factors, including many with bilaterian orthologs, associate with diverse neurosensory structures. Compared to Clytia, the polyp-only hydrozoan Hydra has lost many of the medusa-expressed transcription factors, despite similar overall rates of gene content and sequence evolution. Absence of expression and gene loss among Clytia orthologs of genes patterning the anthozoan aboral pole, secondary axis and endomesoderm support simplification of planulae and polyps in Hydrozoa, including loss of bilateral symmetry. Consequently, although the polyp and planula are generally considered the ancestral cnidarian forms, in Clytia the medusa maximally deploys ancestral cnidarian-bilaterian transcription factor gene complexity.

evolutionary biology↗

bioRxiv2017

Bayesian Selection Of Grammar Productions For The Language Of Thought

Probabilistic proposals of Language of Thoughts (LoTs) can explain learning across different domains as statistical inference over a compositionally structured hypothesis space. While frameworks may differ on how a LoT may be implemented computationally, they all share the property that they are built from a set of atomic symbols and rules by which these symbols can be combined. In this work we show how the set of productions of a LoT grammar can be effectively selected from a broad repertoire of possible productions by an inferential process starting from experimental data. We then test this method in the language of geometry, a specific LoT model (Amalric et al., 2017). Finally, despite the fact of the geometrical LoT not being a universal (i.e. Turing-complete) language, we show an empirical relation between a sequences probability and its complexity consistent with the theoretical relationship for universal languages described by Levins Coding Theorem.

neuroscience↗