Biology subjects

de Souza, N.

Publications and source records attributed to de Souza, N..

15 recordsLinked to original sources

bioRxiv2026

Systematic Evaluation of Feature Representations for Cancer-Associated sORF Prediction in Non-coding RNA

Short open reading frames (sORFs) within non-coding RNAs (ncRNAs) have arisen as a hidden layer of gene regulation, encoding small peptides that represent a new class of cancer regulators with diagnostic and therapeutic potential. However, inferring associations between sORFs to specific cancer types remains challenging and requires computational approaches for accurate prediction. Recently, the CoraL framework introduced the first computational approach for predicting cancer-associated peptides, focusing primarily on model architecture while overlooking how feature extraction strategies influence predictive accuracy. We present a systematic evaluation of machine learning models and feature extraction approaches to predict cancer-associated sORFs across 15 cancer types. We benchmarked seven traditional machine learning algorithms combined with three feature extraction methods: k-mer frequency, Word2Vec embeddings, and genomic language model (gLM)-based embeddings. To our knowledge, this is the first study applying gLM-derived embeddings to the prediction of cancer-associated sORFs in ncRNA. Our results show that traditional machine learning models with appropriate feature extraction outperform the CoraL baseline across all cancer types, achieving up to 10% higher accuracy in some of the 15 evaluated datasets. Interestingly, k-mer features consistently outperformed gLM embeddings without fine-tuning, suggesting that local sequence composition may provide more discriminative information for this task and that pre-trained genomic representations may require task-specific adaptation to fully capture these patterns. Additionally, we observed that the way sequences are tokenized, such as the k-mer length, can affect performance: longer fragments (e.g., k=7) sometimes reduced accuracy for Random Forest but had a smaller effect on MLP. Our findings suggest that appropriate feature engineering can provide greater improvements than increasing model complexity.

bioinformatics↗

bioRxiv2026

Deconvolving the structural heterogeneity of alpha-Synuclein in vitro and in situ

The structural states of proteins in cells and tissues provide important insight into their functional states, but studying protein structures in situ remains challenging. Furthermore, a single protein can adopt multiple conformations in cells, which typically cannot be assessed by most structural approaches. Here we developed a novel approach, based on structural proteomics fingerprints, for the quantitative analysis of the distribution of structural states of a protein in vitro and in situ. We applied it to the Parkinsons disease hallmark protein alpha-synuclein (aSyn), for which various structural states (disordered, helical, oligomeric and amyloid fibrillar, among others) have been characterized in vitro, but for which the in vivo structural states remain hotly debated. We measured structure-specific proteolytic fingerprints from well-characterized aSyn in vitro conformations and used them to quantitatively determine the aSyn conformational composition in samples of interest. We first benchmarked our approach using ground truth datasets of known composition and showed that, during in vitro amyloid fibril formation, we could simultaneously detect a time-dependent decrease in disordered monomeric aSyn, an increase in {beta}-sheet-rich oligomers, and a delayed rise in amyloid fibrils. We then applied the method to complex, biologically relevant samples. In a S. cerevisiae aSyn overex-pression model, aSyn was predominantly helical, with an increased helical fraction accompanying its relocalization from the plasma membrane to cytosolic lipid droplets. This shift was linked to proteome-wide changes in lipid droplet homeostasis and fatty acid and ergosterol metabolism, underscoring the role of lipid metabolism and droplet formation in aSyn biology. Importantly, we also detected helical aSyn in human iPSC-derived cortical neurons, supporting the physiological relevance of this conformation. Finally, neurons differentiated from PD patient-derived iPSCs showed elevated levels of {beta}-sheet-rich aSyn compared to wild-type cells. Our approach allowed the in situ identification and quantification of different structural states of aSyn directly in cell lysates. Since several proteins can adopt multiple, functionally-relevant conformations in cells, our approach should be broadly applicable to in situ, quantitative structural and functional studies of proteins.

neuroscience↗

bioRxiv2026

In-source fragmentation in mass spectrometry-based proteomics: prevalence, impact, and strategies for mitigation

Peptide-level analyses are becoming increasingly popular in mass spectrometry-based proteomics and are being applied, for example, in immunopeptidomics, structural proteomics, and analyses of post-translational modifications. In such analyses, peptides that are not biologically meaningful but instead arise as artifacts prior to mass spectrometry analysis pose the risk of data misinterpretation. Here, we describe an approach based on retention time analysis and precise chromatographic peak matching to identify peptides generated by in-source fragmentation (ISF), which occurs between chromatographic separation of peptide mixtures and the first mass filter of a tandem mass spectrometer (MS). To understand the prevalence and properties of ISF, we generated 13 proteomics datasets and analyzed them along with additional 25 previously published datasets spanning a broad range of sample types, MS, and proteomics approaches including classical bottom-up proteomics, immunopeptidomics, structural proteomics, and phosphoproteomics. We found that, in typical trypsin-digested samples on average 1 % of fully-tryptic peptides and 22 % of semi-tryptic peptides originated from ISF. However, we observed large variations between datasets, and in-source fragments exceeded, in some cases, a third of the total peptide identifications. The extent of ISF was dependent on the peptide sequence, the instrument, method parameters, and sample complexity. Although ISF did not impair relative quantification across samples, it generated peptides that could be misinterpreted qualitatively, inflated peptide identifications, and comprised up to 37 percent of peptides shorter than 9 amino acids in immunopeptidomics datasets. We propose that, for peptide-centric applications, our open-source ISF detection approach be used to re-annotate peptides generated by ISF and remove them to avoid misinterpretation of data. ISF is an increasing concern with improving mass spectrometers, as they enable detection of an ever-increasing number of m/z features, including low abundance features like ISF products. Our work thus addresses a growing issue in proteomics and presents solutions to mitigate the impact of in-source fragment peptides. In the future, improved feature detection algorithms may enable elucidation of new ISF patterns affecting side chains that have been missed so far, which could contribute to explaining the vast space of as-yet unannotated proteomics data.

biochemistry↗

bioRxiv2025

Systematic structural and functional analysis of metal-binding sites in native proteomes

Metal-binding proteins (MBPs) are widespread and critical for many biological processes. Proteomics-based approaches have identified MBPs in various systems but typically cannot determine the location of metal-binding sites or shed light on the functional roles of metals. Thus, a systems-wide map of metal binding sites and an understanding of the consequences of metal binding for protein function are still lacking. Here, we used limited proteolysis-coupled mass spectrometry (LiP-MS) combined with chelator treatment to globally map metal-protein interactions and binding sites in native protein lysates. We applied the pipeline, named Metal-LiP, to Escherichia coli, identified 981 peptides on 297 proteins that exhibited dose responsiveness to chelator treatment, and quantified the relative metal-binding affinity. These proteins are strongly enriched for known or predicted MBPs, based on our novel compiled ground truth dataset of previously annotated (n=1013) and predicted (n=850) MBPs. Further, chelator-responsive peptides mapped at or near annotated metal-binding sites in 3D protein structures, showing that Metal-LiP can identify metal-binding sites, and provided novel experimental evidence for previously predicted sites. Notably, Metal-LiP identified 225 novel candidate MBPs and we validated five of these novel MBPs (GatZ, AhpC, GrxB, MsrC, TrhO) using inductively coupled plasma-mass spectrometry to detect metal binding in purified proteins. Finally, we used orthogonal approaches to probe the broad functional consequences of metal binding. Using a combination of Metal-LiP, thermal proteome profiling, size-exclusion chromatography coupled to mass spectrometry, and structural analyses, we identified cases where metal binding affects protein stability, modulates the assembly states of protein complexes, or is likely to affect enzymatic activity. We thereby propose functional roles for metal binding in 98 of our Metal-LiP hits and in 225 proteins in total. In summary, we describe an approach to identify bona fide MBPs with binding site-level resolution across the proteome. We have combined it with other methods to identify 424 candidate MBPs in the E coli proteome, to map known and novel candidate metal-binding sites for 297 of these proteins, and to suggest functional roles of metal binding.

systems biology↗

bioRxiv2025

Structure-function relationship of alpha-synuclein fibrillar polymorphs derived from distinct synucleinopathies

The aggregation of the protein alpha-synuclein (Syn) is a common feature of multiple neurodegenerative diseases collectively called synucleinopathies, for which the pathobiology is not well understood. The different phenotypic characteristics of the synucleinopathies Parkinsons disease (PD), Dementia with Lewy Bodies (DLB) and Multiple System Atrophy (MSA) have been proposed to originate from the distinct structures adopted by Syn in its amyloid forms. Here, using covalent labeling and limited proteolysis coupled to mass spectrometry (LiP-MS) in vitro and in situ within neuronal cells and directly in native patient brain homogenates, we show that pathogenic Syn from distinct synucleinopathies (PD, DLB and MSA) are structurally different. Further, we found that fibrillar structural differences are associated with different fibril interactomes and neuronal responses. We discovered disease-specific ubiquitination patterns and turnover profiles for pathogenic Syn species, detected molecular pathways responding specifically to the uptake of different Syn fibrillar polymorphs, and identified a subset of the involved proteins as candidate direct interactors of Syn. In particular, components of the Ubiquitin-proteasomal System (UPS), including E3 ubiquitin ligases, chaperones, and Deubiquitinating proteins, showed disease/polymorph-specific interaction patterns, possibly accounting for different resistance of patient-derived Syn fibrils to degradation. Genetic modulation with CRISPR-based tools showed that members of the UPS degradation pathway (three E3 ligases: UBE3A, TRIM25, HUWE1 and the AAA+ ATPase VCP) reduced Syn inclusions, in a strain-specific manner. LiP-MS also identified sets of proteins with altered protease susceptibility in postmortem brain homogenates of PD, DLB, and MSA patients. These sets were largely disease-specific and included proteins altered in cells treated with fibrils derived from patients with the matching disease. Our findings provide insight into cellular processes involved in the accumulation and turnover of Syn pathogenic aggregates in PD, DLB and MSA in a disease specific manner and constitutes a resource of potential novel drug targets in these synucleinopathies.

systems biology↗

bioRxiv2025

Imaging mass cytometry reveals early β-cell dysfunction and changes in immune signatures during type 1 diabetes progression in human pancreata

The natural history and pathogenesis of type 1 diabetes, particularly during the autoantibody- positive stages preceding clinical onset, are not well understood, in part, due to limited availability of human pancreatic samples. Here, we studied 88 organ donors, including 28 single autoantibody-positive and 10 multiple autoantibody-positive donors, by imaging mass cytometry. Approximately 10,000 islets and 16 million single-cells were spatially analyzed using 79 antibodies revealing both {beta}-cell states and the islet-immune interface. We identified IAPP loss from {beta}-cells as an indicator of pre-clinical disease. Alterations in Interferon signatures and downregulation across lineage and functional markers, including markers of endoplasmic reticulum stress, were characteristic of recent-onset disease. Further, in single autoantibody- positive donors, we identified pro-inflammatory myeloid cells and PD1+ memory CD4+ T cells, and in multiple autoantibody-positive samples, found islet-specific and exhausted-like ebector CD8+ T cells. Multiple immune cell subtypes were associated with young age, disease severity and insulitis. This dataset is a major step toward creation of a multi-modal type 1 diabetes disease atlas that will be useful for identifying potential drug targets and association of disease features with clinical co-variates and trial outcomes.

immunology↗

bioRxiv2025

Limited proteolysis-coupled mass spectrometry captures proteome-wide protein structural alterations and biomolecular condensation in living cells

The function of a protein is determined by its structure, which may change dynamically in response to post-translational modifications, interaction with other molecules, or environmental factors like temperature. Limited proteolysis-coupled mass spectrometry (LiP-MS) captures such structural alterations on a proteome-wide scale via the detection of altered protease susceptibility patterns of proteins. However, this technique has so far required cell lysis, which exposes proteins to non-native conditions and can disrupt labile interactions such as those occurring within biomolecular condensates. To study protein structures directly within cells, we developed in-cell LiP-MS. We optimized conditions for introduction of proteinase K into human cells using electroporation and validated that intracellular cleavage occurs. In-cell LiP-MS captured not only the known binding of rapamycin to FKBP1A within the cell, but also downstream effects of pathway activation. Moreover, it detected global protein structural alterations upon sodium arsenite treatment and captured the structural dynamics of hundreds of proteins from biomolecular condensates with peptide level resolution and within live human cells. We detected known and novel structural alterations of proteins from stress granules as well as from nuclear speckles and validated alteration of nuclear speckles by fluorescence microscopy. Our dataset provides a resource describing the structural changes of human proteins in response to cellular stress and pinpoints structurally altered regions. Further, comparison of LiP-based structural fingerprints before and after cell lysis revealed which human proteins are susceptible to structural change upon cell lysis, therefore guiding the design of future experiments requiring native protein structures.

systems biology↗

bioRxiv2023

An approach to characterize mechanisms of action of anti-amyloidogenic compounds in vitro and in situ

Aggregation of amyloidogenic proteins is associated with neurodegenerative disease and its modulation is a focus of drug development efforts. However, the physicochemical properties and structural heterogeneity of amyloidogenic proteins hinder the mechanistic understanding of anti- amyloidogenic compounds. Further, modes of interaction with amyloidogenic proteins are often probed in vitro using purified protein samples, even though these models may not capture in vivo protein structures and do not enable identification of off-target effects. We have developed a modular structural proteomic pipeline based on limited proteolysis coupled to mass spectrometry (LiP-MS) with improved, amino acid level-resolution, to probe the mechanism of action of anti- amyloidogenic compounds. We demonstrate our approach by analysing the interactions of six known or putative anti-amyloidogenic compounds and the amyloid binder Thioflavin T (ThT) with different structural forms of the amyloidogenic Parkinsons disease (PD) protein -Synuclein. Our approach enables determination of putative interaction sites, identification of whether interactions are covalent or non-covalent, and crucially, can probe for interactions of compounds with physiological structures of -Synuclein in complex cell and tissue extracts and identify off-targets. In vitro analyses with our pipeline showed that the green tea polyphenol EGCG induces an N- and C-terminus- dependent compaction of the unstructured -Synuclein monomer, detected preferential interactions of ThT with the N-terminus of -Synuclein fibrils compared to the amyloid core, and showed that the most potent inhibitors of aggregation in our study (EGCG, baicalein and AC Immune compound #2) induced similar non-fibrillar end structures despite different interactions with -Synuclein monomers. Importantly, in mammalian cell lysates, -Synuclein was either a low-affinity target (for EGCG and Baicalein) or did not show evidence of compound interaction (for ThT and doxycycline) in our experimental conditions, despite both monomeric and fibrillar forms interacting with these compounds in vitro. For EGCG, we validated this result in postmortem brain homogenates from PD patients. These in situ analyses identified many additional putative cellular targets of Doxycycline, EGCG, Baicalein and ThT, suggesting that their effects in cellular or animal models of neurodegeneration are likely due to interactions with proteins other than -Synuclein and showing that anti-amyloidogenic compounds should be analyzed in situ as well as in vitro. Our modular pipeline will enable in situ screening of drugs and PET tracers for amyloid aggregates of interest as well as detailed mechanistic studies of compound action in vitro.

systems biology↗

bioRxiv2023

Systematic identification of structure-specific protein-protein interactions

Protein-protein interactions (PPIs) mediate numerous essential functions and regulatory events in living organisms. The physical interactome of a protein can be abnormally altered in response to external and internal cues, thus modulating cell physiology and contributing to human disease. In particular, neurodegenerative diseases due to the accumulation of aberrantly folded and aggregated proteins may lead to alterations in protein interactomes. Identifying changes in the interactomes of normal and disease states of proteins could help to understand molecular disease mechanisms, but current interactomics methods are limited in the ability to pinpoint structure-specific PPIs and their interaction interfaces on a proteome-wide scale. Here, we adapted limited proteolysis-mass spectrometry (LiP-MS) to systematically identify putative structure-specific PPIs by probing protein structural alterations within cellular extracts upon treatment with specific structural states of a given protein. We demonstrate the feasibility of our method to detect well-characterized PPIs, including antibody-target protein interactions and interactions with membrane proteins, and show that it pinpoints PPI interfaces. We then applied the LiP-MS approach to study the structure-specific interactors of the Parkinsons disease hallmark protein alpha-synuclein (aSyn). We identified several previously known interactors of both aSyn monomer and amyloid fibrils and provide a resource of novel putative structure-specific interactors for further studies. This approach is applicable to identify structure-specific interactomes of any protein, including posttranslationally modified and unmodified, or metabolite-bound and unbound structural states of proteins.

systems biology↗

bioRxiv2022

A Cancer-Associated Fibroblast Classification Framework for Single-Cell Data

Cancer-associated fibroblasts (CAFs) are a diverse cell population within the tumour microenvironment, where they have critical effects on tumour evolution and patient prognosis. To define CAF phenotypes, we analysed a single-cell RNA sequencing (scRNA-seq) dataset of over 16,000 stromal cells from tumours of 14 breast cancer patients, based on which we defined and functionally annotated nine CAF phenotypes and one class of pericytes. We validated this classification system in four additional cancer types and used highly multiplexed imaging mass cytometry on matched breast cancer samples to confirm our defined CAF phenotypes at the protein level and to analyse their spatial distribution within tumours. This general CAF classification scheme will allow comparison of CAF phenotypes across studies, facilitate analysis of their functional roles, and potentially guide development of new treatment strategies in the future.

cancer biology↗

bioRxiv2021

Mass Cytometric and Transcriptomic Profiling of Epithelial-Mesenchymal Transitions in Human Mammary Cell Lines

Epithelial-mesenchymal transition (EMT) equips breast cancer cells for metastasis and treatment resistance. Inhibition and elimination of EMT-undergoing cells are therefore promising therapy approaches. However, detecting EMT-undergoing cells is challenging due to the intrinsic heterogeneity of cancer cells and the phenotypic diversity of EMT programs. Here, we profiled EMT transition phenotypes in four non-cancerous human mammary epithelial cell lines using a FACS surface marker screen, RNA sequencing, and mass cytometry. EMT was induced in the HMLE and MCF10A cell lines and in the HMLE-Twist-ER and HMLE-Snail-ER cell lines by chronic exposure to TGF{beta}1 or 4-hydroxytamoxifen, respectively. We observed a spectrum of EMT transition phenotypes in each cell line and the spectrum varied across the time course. Our data provide multiparametric insights at single-cell level into the phenotypic diversity of EMT at different time points and in four human cellular models. These insights are valuable to better understand the complexity of EMT, to compare EMT transitions between the cellular models used herein, and for the design of EMT time course experiments. O_TBL View this table: org.highwire.dtl.DTLVardef@10d2b87org.highwire.dtl.DTLVardef@1a4cc67org.highwire.dtl.DTLVardef@67bc95org.highwire.dtl.DTLVardef@8df40dorg.highwire.dtl.DTLVardef@1919e7c_HPS_FORMAT_FIGEXP M_TBL C_TBL Mendeley Data: DOI: 10.17632/pt3gmyk5r2.1 ArrayExpress Data: Accession number E-MTAB-9365

cell biology↗

bioRxiv2020

A distinct innate immune signature marks progression from mild to severe COVID-19

Coronavirus disease 2019 (COVID-19) manifests with a range of severities, but immune signatures of mild and severe disease are still not fully understood. Excessive inflammation has been postulated to be a major factor in the pathogenesis of severe COVID-19 and innate immune mechanisms are likely to be central in the inflammatory response. We used 40-plex mass cytometry and targeted serum proteomics to profile innate immune cell populations from peripheral blood of patients with mild or severe COVID-19 and healthy controls. Sampling at different stages of COVID-19 allowed us to reconstruct a pseudo-temporal trajectory of the innate immune response. Despite the expected patient heterogeneity, we identified consistent changes during the course of the infection. A rapid and early surge of CD169+ monocytes associated with an IFN{gamma}+MCP-2+ signature quickly followed symptom onset; at symptom onset, patients with mild and severe COVID-19 had a similar signature, but over the course of the disease, the differences between patients with mild and severe disease increased. Later in the disease course, we observed a more pronounced re-appearance of intermediate/non-classical monocytes and mounting systemic CCL3 and CCL4 levels in patients with severe disease. Our data provide new insights into the dynamic nature of the early inflammatory response to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection and identifies sustained pathological innate immune responses as a likely key mechanism in severe COVID-19, further supporting investigation of targeted anti-inflammatory interventions in severe COVID-19.

immunology↗

bioRxiv2020

A quantitative analysis of the interplay of environment, neighborhood and cell state in 3D spheroids

1Cells react to their microenvironment by integrating external stimuli into phenotypic decisions via an intracellular signaling network. Even cells with deregulated signaling can adapt to their environment. To analyze the interplay of environment, neighborhood, and cell state on phenotypic variability, we developed an experimental approach that enables multiplexed mass cytometric imaging to analyze up to 240 pooled spheroid microtissues. This system allowed us to quantify the contributions of environment, neighborhood, and intracellular state to phenotypic variability in spheroid cells. A linear model explained on average more than half of the variability of 34 markers across four cell lines and six growth conditions. We found that the contributions of cell-intrinsic and environmental factors are hierarchically interdependent. By overexpression of 51 signaling protein constructs in subsets of cells, we identified proteins that have cell-intrinsic and extrinsic effects, exemplifying how cell states depend on the cellular neighborhood in spheroid culture. Our study deconvolves factors influencing cellular phenotype in a 3D tissue and provides a scalable experimental system, analytical principles, and rich multiplexed imaging datasets for future studies.

systems biology↗

bioRxiv2020

Deciphering the Signaling Network Landscape of Breast Cancer Improves Drug Sensitivity Prediction

Although genetic and epigenetic abnormalities in breast cancer have been extensively studied, it remains difficult to identify those patients who will respond to particular therapies. This is due in part to our lack of understanding of how the variability of cellular signaling affects drug sensitivity. Here, we used mass cytometry to characterize the single-cell signaling landscapes of 62 breast cancer cell lines and five lines from healthy tissue. We quantified 34 markers in each cell line upon stimulation by the growth factor EGF in the presence or absence of five kinase inhibitors. These data - on more than 80 million single cells from 4,000 conditions - were used to fit mechanistic signaling network models that provide unprecedented insights into the biological principles of how cancer cells process information. Our dynamic single-cell-based models more accurately predicted drug sensitivity than static bulk measurements for drugs targeting the PI3K-MTOR signaling pathway. Finally, we identified genomic features associated with drug sensitivity by using signaling phenotypes as proxies, including a missense mutation in DDIT3 predictive of PI3K-inhibition sensitivity. This provides proof of principle that single-cell measurements and modeling could inform matching of patients with appropriate treatments in the future. One-linerSingle-cell proteomics coupled to perturbations improves accuracy of breast tumor drug sensitivity predictions and reveals mechanisms of sensitivity and resistance. HIGHLIGHTSO_LIMass cytometry study of signaling responses of 62 breast cancer cell lines and five lines from healthy tissue to EGF stimulation with or without perturbation with five kinase inhibitors. C_LIO_LISingle-cell signaling features and mechanistic signaling network models predicted drug sensitivity. C_LIO_LIMechanistic signaling network models deepen the understanding of drug resistance and sensitivity mechanisms. C_LIO_LIWe identify drug sensitivity-predictive genomic features via proxy signaling phenotypes. C_LI

systems biology↗

bioRxiv2019

LiP-Quant, an automated chemoproteomic approach to identify drug targets in complex proteomes

Chemoproteomics is a key technology to characterize the mode of action of drugs, as it directly identifies the protein targets of bioactive compounds and aids in developing optimized small-molecule compounds. Current unbiased approaches cannot directly pinpoint the interaction surfaces between ligands and protein targets. To address his limitation we have developed a new drug target deconvolution approach based on limited proteolysis coupled with mass spectrometry that works across species including human cells (LiP-Quant). LiP-Quant features an automated data analysis pipeline and peptide-level resolution for the identification of any small-molecule binding sites, Here we demonstrate drug target identification by LiP-Quant across compound classes, including compounds targeting kinases and phosphatases. We demonstrate that LiP-Quant estimates the half maximal effective concentration (EC50) of compound binding sites in whole cell lysates. LiP-Quant identifies targets of both selective and promiscuous drugs and correctly discriminates drug binding to homologous proteins. We finally show that the LiP-Quant technology identifies targets of a novel research compound of biotechnological interest.

systems biology↗