Biology subjects

Sarti, E.

Publications and source records attributed to Sarti, E..

12 recordsLinked to original sources

bioRxiv2026

Evaluating transformer-based models for structural characterization of orphan proteins

MotivationTransformer-based models (TBMs) are state-of-the-art deep learning architectures that predict protein structural features with high accuracy. Despite methodological differences, they all rely on large protein sequence datasets structured by homology, as homologous proteins typically share similar structures. However, 5-30% of eukaryotic proteomes consist of orphan proteins--sequences without detectable similarity to known families. Although they may share structural traits with characterized proteins, their lack of homology makes them and ideal dataset for evaluating TBM generalization beyond familiar sequence space. ResultsWe compared predictions from several widely used TBM architectures on an expert-curated set of orphan proteins from the Meloidogyne genus. None of these proteins has an experimentally determined structure. To assess model performance, we conducted consistency analyses, comparing predicted features with those observed in sets of known homologous proteins and across models. Multiple sequence alignment-based approaches such as AlphaFold2 performed poorly on orphan proteins, as did single-sequence or embedding-based language models including ESMFold, OmegaFold, and ProtT5. This limited performance cannot be fully attributed to intrinsic disorder, as confirmed by independent non-TBM disorder predictors. While accurate tertiary structure prediction remains out of reach, secondary structure is more reliably captured: predictors share about 70% of secondary structure elements on average, regardless of global fold similarity, and these elements are consistently identified by dedicated secondary structure tools. AvailabilityAll data and analysis scripts are available at https://doi.org/10.5281/zenodo.18788931 Contactedoardo.sarti@inria.fr

bioinformatics↗

bioRxiv2025

Identification, evolutionary history and characteristics of orphan genes in root-knot nematodes

Orphan genes, lacking detectable homologs in other species, are common across eukaryotic genomes and can arise through divergence of existing genes or de novo from non-coding regions. Here we identified and characterized orphan genes in eight root-knot nematode species (genus Meloidogyne), the most destructive plant-parasitic nematodes. For that, we used comparative genomics across 85 nematodes, ancestral sequence reconstruction, synteny analyses, and multi-omics data. We found that around 16% of Meloidogyne genes are genus-specific transcribed orphan genes, with about 20% resulting from high divergence and 18% emerging de novo, often in transposon-rich genomic regions. Transcriptomic, translatomic, and proteomic evidence confirmed expression and translation of many orphan genes, which tend to encode shorter, secreted proteins. Notably, orphan genes are preferentially expressed during the infective juvenile stage and contribute substantially to the arsenal of nematode parasitism effectors. These findings highlight orphan genes as a significant and dynamic component of Meloidogyne genomes, potentially underpinning their parasitic adaptation and success.

bioinformatics↗

bioRxiv2025

Fold or flop: quality assessment of AlphaFold predictions on whole proteomes

MotivationReliability of AlphaFold2 predictions is mainly assessed using the predicted Local Distance Difference Test (pLDDT). For model organisms, 30-40% of residues fall into the low-confidence pLDDT range. Moreover, pLDDT sometimes fails to flag physically implausible structures. This raises two questions: can more robust reliability indicators be identified, and do unreliable predictions share common structural or biophysical features? ResultsWe characterize protein structures through histograms of per-residue neighbor counts, and use the Wasserstein principal component analysis to define the arity map, and lightweight and informative 2D embedding of proteins in a dataset. Using AlphaFold-DB, we show that the arity map reveals three structurally and biophysically distinct populations (well-folded proteins, intrinsically disordered proteins, and physically implausible predictions). We also use our packing based encoding at the residue level to define abstraqt (Arity-Based STRuctural Arrangement Quality assessmenT), a per-residue scoring function complementing the pLDDT, assigning low scores to hallucinated helices and distorted beta strands while correctly scoring native like predictions. AvailabilityThe code to compute arity maps is available within the Structural Bioinformatics Library. See: AlphaFold analysis, and also Documentation, Applications, Installation guide. The code and data for rerunning analyses are made available from AlphaFold analysis, and Data.

bioinformatics↗

bioRxiv2025

Characterizing the fragmentation of AlphaFold predictions

The Nobel prize winning program AlphaFold2 computes plausible structures of (well) folded proteins. The main quality assessment is based on the predicted Local Distance Difference Test (pLDDT), a per amino acid confidence score. To enhance quality assessment, we provide novel quantitative measures to identify coherent amino acid (a.a.) stretches along the sequence in terms of pLDDT values. These constructions, grounded in standard techniques from topological data analysis and combinatorics, provide a canonical framework for identifying regions along the protein backbone and analyzing their properties, such as their propensity for disorder and their consistency with a null model. The outcome of our analysis can readily be used to select reliable regions/domains within proteins whose pLDDT values span the entire pLDDT range.

bioinformatics↗

bioRxiv2025

c-Myc inhibits macrophage antimycobacterial response in Mycobacterium tuberculosis infection

Mycobacterium tuberculosis (MTB) is a major global cause of mortality worldwide, responsible for over a million deaths annually. Despite this burden, natural immunity prevents disease in more than 90% of exposed individuals. Previous studies have identified interferon-gamma (IFN-{gamma}) as a key regulator of innate immune defense against MTB. Here, we investigate the impact of IFN-{gamma} timing on macrophage-mediated control of MTB infection. We demonstrate that IFN-{gamma} exposure before infection enhances macrophage antibacterial activity, whereas post-infection exposure does not. Further investigation into this phenotype revealed a strong association between c-Myc signaling and macrophage function in MTB control, as identified using unbiased in vitro systems approaches. Given the challenge of perturbing c-Myc in primary cells, we developed a lentiviral system for c-Myc inhibition and overexpression. Using a tetracycline-inducible Omomyc system - a small peptide inhibitor of c-Myc - we show that c-Myc inhibition promotes a pro-inflammatory macrophage phenotype with enhanced antimycobacterial activity. Mechanistically, c-Myc inhibition induces metabolic reprogramming via increased mTORC1 activity, leading to upregulated inducible nitric oxide synthase and improved bacterial control. In vivo analyses, including murine models and human clinical histopathology, reveal a strong correlation between c-Myc expression and MTB persistence, as well as active tuberculosis (TB), suggesting a role for c-Myc in immune evasion. These findings reveal c-Myc as a potential mediator of immune privilege in MTB infection and highlight its role as a promising target for novel TB therapies aimed at enhancing macrophage function.

immunology↗

bioRxiv2024

AlphaFold predictions on whole genomes at a glance

For model organisms, AlphaFold predictions show that 30% to 40% of amino acids have a (very) low pLDDT confidence score. This observation, combined with the methods high complexity, commands a systematic analysis of AlphaFold predictions on whole genomes. Consequently, using whole-genome predictions, we provide a coherent analysis on packing properties, pLDDT values, and their relationship with intrinsically disordered regions (IDRs). Our contributions are of two kinds. First, we introduce simple and explainable geometrical and topological statistics characterizing predictions. Second, we investigate four key biophysical and biological questions: (i) the clustering of AlphaFold predictions on whole genomes, (ii) the identification of high/low quality predicted domains, (iii) false positive/negative AlphaFold predictions with respect to IDRs, and (iv) the fragmentation of the polypeptide chain in terms of pLDDT values. Altogether, our analysis provide novel insights into AlphaFold predictions across whole genomes, further enhancing the confidence assessment of the models.

bioinformatics↗

bioRxiv2024

Simpler protein domain identification using spectral clustering

The decomposition of a biomolecular complex into domains is an important step to investigate biological functions and ease structure determination. A successful approach to do so is the SPECTRUS algorithm, which provides a segmentation based on spectral clustering applied to a graph coding interatomic fluctuations derived from an elastic network model. We present SPECTRALDOM, which makes three straightforward and useful additions to SPECTRUS. For single structures, we show that high quality partitionings can be obtained from a graph Laplacian derived from pairwise interactions-without normal modes. For sets of homologous structures, we introduce a Multiple Sequence Alignment mode, exploiting both the sequence based information (MSA) and the geometric information embodied in experimental structures. Finally, we propose to analyse the clusters/- domains delivered using the so-called D-family-matching algorithm, which establishes a correspondence between domains yielded by two decompositions, and can be used to handle fragmentation issues. Our domains compare favorably to those of the original SPECTRUS, and those of the deep learning based method Chainsaw. Using two complex cases, we show in particular that SPECTRALDOM is the only method handling complex conformational changes involving several sub-domains. Finally, a comparison of SPECTRALDOM and Chainsaw on the manually curated domain classification ECOD as a reference shows that high quality domains are obtained without using any evolutionary related piece of information. SPECTRALDOM is provided in the Structural Bioinformatics Library, see http://sbl.inria.fr and https://sbl.inria.fr/doc/Spectral_domain_explorer-user-manual.html.

bioinformatics↗

bioRxiv2021

Crystal structure of chloroplast fructose-1,6-bisphosphate aldolase from the green alga Chlamydomonas reinhardtii

The Calvin-Benson cycle fixes carbon dioxide into organic triosephosphates through the collective action of eleven conserved enzymes. Regeneration of ribulose-1,5-bisphosphate, the substrate of Rubisco-mediated carboxylation, requires two lyase reactions catalyzed by fructose-1,6-bisphosphate aldolase (FBA). While cytoplasmic FBA has been extensively studied in non-photosynthetic organisms, functional and structural details are limited for chloroplast FBA encoded by oxygenic phototrophs. Here we determined the crystal structure of plastidial FBA from the unicellular green alga Chlamydomonas reinhardtii (Cr). We confirm that CrFBA folds as a TIM barrel, describe its catalytic pocket and homo-tetrameric state. Multiple sequence profiling classified the photosynthetic paralogs of FBA in a distinct group from non-photosynthetic paralogs. We mapped the sites of thiol- and phospho-based post-translational modifications known from photosynthetic organisms and predict their effects on enzyme catalysis.

biochemistry↗

bioRxiv2020

Refining pairwise sequence alignments of membrane proteins by the incorporation of anchors

The alignment of primary sequences is a fundamental step in the analysis of protein structure, function, and evolution. Integral membrane proteins pose a significant challenge for such sequence alignment approaches, because their evolutionary relationships can be very remote, and because a high content of hydrophobic amino acids reduces their complexity. Frequently, biochemical or biophysical data is available that informs the optimum alignment, for example, indicating specific positions that share common functional or structural roles. Currently, if those positions are not correctly aligned by a standard pairwise alignment procedure, the incorporation of such information into the alignment is typically addressed in an ad hoc manner, with manual adjustments. However, such modifications are problematic because they reduce the robustness and reproducibility of the alignment. An alternative approach is the use of restraints, or anchors, to incorporate such position-matching explicitly during alignment. Here we introduce position anchoring in the alignment tool AlignMe as an aid to pairwise sequence alignment of membrane proteins. Applying this approach to realistic scenarios involving distantly-related and low complexity sequences, we illustrate how the addition of even a single anchor can dramatically improve the accuracy of the alignments, while maintaining the reproducibility and rigor of the overall alignment.

bioinformatics↗

bioRxiv2019

FilterDCA: interpretable supervised contact prediction using inter-domain coevolution

Predicting three-dimensional protein structure and assembling protein complexes using sequence information belongs to the most prominent tasks in computational biology. Recently substantial progress has been obtained in the case of single proteins using a combination of unsupervised coevolutionary sequence analysis with structurally supervised deep learning. While reaching impressive accuracies in predicting residue-residue contacts, deep learning has a number of disadvantages. The need for large structural training sets limits the applicability to multi-protein complexes; and their deep architecture makes the interpretability of the convolutional neural networks intrinsically hard. Here we introduce FilterDCA, a simpler supervised predictor for inter-domain and inter-protein contacts. It is based on the fact that contact maps of proteins show typical contact patterns, which results from secondary structure and are reflected by patterns in coevolutionary analysis. We explicitly integrate averaged contacts patterns with coevolutionary scores derived by Direct Coupling Analysis, reaching results comparable to more complex deep-learning approaches, while remaining fully transparent and interpretable. The FilterDCA code is available at http://gitlab.lcqb.upmc.fr/muscat/FilterDCA. Author summaryThe de novo prediction of tertiary and quaternary protein structures has recently seen important advances, by combining unsupervised, purely sequence-based coevolutionary analyses with structure-based supervision using deep learning for contact-map prediction. While showing impressive performance, deep-learning methods require large training sets and pose severe obstacles for their interpretability. Here we construct a simple, transparent and therefore fully interpretable inter-domain contact predictor, which uses the results of coevolutionary Direct Coupling Analysis in combination with explicitly constructed filters reflecting typical contact patterns in a training set of known protein structures, and which improves the accuracy of predicted contacts significantly. Our approach thereby sheds light on the question how contact information is encoded in coevolutionary signals.

bioinformatics↗

bioRxiv2019

MemSTATS: A Benchmark Set of Membrane Protein Symmetries and Pseudo-Symmetries

In membrane proteins, symmetry and pseudo-symmetry often have functional or evolutionary implications. However, available symmetry detection methods have not been tested systematically on this class of proteins due to the lack of an appropriate benchmark set. Here we present MemSTATS, a publicly-available benchmark set of both quaternary and internal symmetries in membrane protein structures described in terms of order, repeated elements, and orientation of the axis with respect to the membrane plane. Moreover, using MemSTATS, we compare the performance of four widely-used symmetry detection algorithms and highlight specific challenges and areas for improvement in the future.

bioinformatics↗

bioRxiv2018

EncoMPASS: an Encyclopedia of Membrane Proteins Analyzed by Structure and Symmetry

Protein structure determination and prediction, active site detection, and protein sequence alignment techniques all exploit information about protein structure and structural relationships. For membrane proteins, however, there is no agreement among available online tools for highlighting and mapping such structural similarities. Moreover, no available resource provides a systematic overview of quaternary and internal symmetries, and their orientation with respect to the membrane, despite the fact that these properties can provide key insights into membrane protein function. To address these issues, we created the Encyclopedia of Membrane Proteins Analyzed by Structure and Symmetry (EncoMPASS), a database for relating integral membrane proteins of known structure from the points of view of sequence, structure, and symmetry. EncoMPASS is accessible at https://encompass.ninds.nih.gov and its contents can be easily downloaded. This allows the user not only to focus on specific systems, but also to study general properties of the structure and evolution of membrane proteins.\n\nHighlights-EncoMPASS relates and analyzes known structures of membrane proteins\n-Structure and sequence similarity is assessed through alignments and topology considerations, not clustering\n-Symmetry is detected based on CE-Symm and SymD using a multi-step procedure

bioinformatics↗