Biology subjects

Silveira, S.

Publications and source records attributed to Silveira, S..

5 recordsLinked to original sources

bioRxiv2026

Mitigating Class Imbalance in Graph-Structured Data via Hierarchical Learning: Insights from Protein Binding Site Prediction

Learning from imbalanced data remains a major challenge for graph neural networks (GNNs), as minority nodes are not only rare but also structurally marginalized within the graph. We address this issue with CLARA, a hierarchical learning framework that decomposes node classification into two stages: a coarse subgraph-level classifier that selects regions likely to contain minority instances, followed by a fine-grained node-level predictor within these regions. This design improves sensitivity while maintaining scalability, filtering out irrelevant areas and focusing learning on topologically meaningful neighborhoods. Experiments on benchmark graph datasets demonstrate substantial gains over established imbalance-handling methods, with CLARA reaching an F1-score of 88.3\%. The same strategy achieves significant improvements in protein-ligand binding site prediction, underscoring its broad and consistent effectiveness across both biological and general graph learning tasks.

bioinformatics↗

bioRxiv2025

PERSEUcpp: A machine learning strategy to predictcell-penetrating peptides and their uptake efficiency

MotivationCell-penetrating peptides (CPPs) are promising tools for transporting therapeutic molecules into cells without damaging the cellular membrane. These peptides serve as efficient drug delivery systems, capable of carrying diverse biologically active substances while exhibiting low cytotoxicity compared to non-native molecules. However, identifying CPPs through experimental methods is expensive and time-consuming, making computational strategies an attractive alternative due to their cost-effectiveness and scalability. ResultsThis study introduces PERSEUcpp, a machine learning strategy designed to identify CPPs. Based on descriptors including physicochemical and structural properties as well as atomic composition, our strategy employs the Extremely Randomized Trees to predict CPPs and their uptake efficiency. The first stage was developed using a balanced dataset of 967 CPPs and non-CPPs, applying a 10-fold cross-validation scheme. Two independent datasets were utilized for validation. The CPP predictor achieved superior results compared to state-of-the-art methods, with MCC 0.854, Recall 0.860, and AUC 0.984. The second stage, focused on efficiency prediction, was trained on a balanced dataset of 140 CPPs and non-CPPs, also using a 10-fold cross-validation scheme, and validated with an independent dataset. The efficiency predictor achieved competitive results, with Recall 0.761 and AUC 0.690. PERSEUcpp is interpretable, offering insights into the key features enabling peptides to penetrate cells effectively. We anticipate that PERSEUcpp will be a valuable tool for advancing the design and application of CPPs in drug delivery and biomedical research. Availability and ImplementationAll the source code and datasets are available at GitHub: https://github.com/goalmeida05/PERSEU.

bioinformatics↗

bioRxiv2025

BENDER DB: a database of protein binding sites across neglected disease proteomes

Identifying binding sites is crucial for expanding our knowledge of various biological processes, supporting drug discovery, and repositioning strategies, particularly in the early research phases of target hopping. This is especially important for neglected diseases, which predominantly impact vulnerable populations in developing regions. To address this, we created BENDER DB, a database designed to map predicted protein binding sites within the proteomes of pathogens associated with neglected diseases. Utilizing AlphaFold-predicted structures, BENDER DB integrates results from five leading binding site prediction tools, resulting in over one million binding sites across over 100,000 proteins from 10 different proteomes. BENDER DB offers unique features, such as integrating multiple predictors, interactive visualization tools, and comprehensive graphical representations, allowing for a detailed visual comparative analysis of different prediction methods and the predicted binding sites. We combined the computational approaches to offer a consensus of binding site predictions, leveraging the strengths of multiple techniques. Additionally, we introduce BENDER AI, a meta-predictor that combines the outputs of the predictors to provide a unified binding site classification. BENDER AI outperformed all individual methods in 4 out of 5 evaluated metrics. Specifically, it achieved an MCC of 0.64 and an AUC of 0.89, surpassing the highest MCC of 0.63 and the highest AUC of 0.84 obtained by individual predictors, demonstrating strong performance and effectiveness in supporting binding site prediction. By consolidating these results, BENDER DB enables detailed comparative analyses of prediction methods and integrates interactive visualization tools to provide users with an intuitive platform for binding site exploration. The database aims to accelerate drug discovery efforts, particularly in underdeveloped regions, by providing a robust and detailed resource for studying protein binding sites. BENDER DB is available at https://benderdb.ufv.br.

bioinformatics↗

bioRxiv2020

Mapping Cognitive Brain Functions at Scale

A fundamental set of cognitive abilities enable humans to efficiently process goal-relevant information, suppress irrelevant distractions, maintain information in working memory, and act flexibly in different behavioral contexts. Yet, studies of human cognition and their underlying neural mechanisms usually evaluate these cognitive constructs in silos, instead of comprehensively in-tandem within the same individual. Here, we developed a scalable, mobile platform, "BrainE" (short for Brain Engagement), to rapidly assay several essential aspects of cognition simultaneous with wireless electroencephalography (EEG) recordings. Using BrainE, we rapidly assessed five aspects of cognition including (1) selective attention, (2) response inhibition, (3) working memory, (4) flanker interference and (5) emotion interference processing, in 102 healthy young adults. We evaluated stimulus encoding in all tasks using the EEG neural recordings, and isolated the cortical sources of the spectrotemporal EEG dynamics. Additionally, we used BrainE in a two-visit study in 24 young adults to investigate the reliability of the neuro-cognitive data as well as its plasticity to transcranial magnetic stimulation (TMS). We found that stimulus encoding on multiple cognitive tasks could be rapidly assessed, identifying common as well as distinct task processes in both sensory and cognitive control brain regions. Event related synchronization (ERS) in the theta (3-7 Hz) and alpha (8-12 Hz) frequencies as well as event related desynchronization (ERD) in the beta frequencies (13-30 Hz) were distinctly observed in each task. The observed ERS/ERD effects were overall anticorrelated. The two-visit study confirmed high test-retest reliability for both cognitive and neural data, and neural responses showed specific TMS protocol driven modulation. We also show that the global cognitive neural responses are sensitive to mental health symptom self-reports. This first study with the BrainE platform showcases its utility in studying neuro-cognitive dynamics in a rapid and scalable fashion. HighlightsO_LIRapid and scalable EEG recordings reveal common and distinct cortical activations across five core cognitive tasks. C_LIO_LIData acquired across visits one-week-apart show high test-retest reliability for both cognitive and neural measurements. C_LIO_LIEvoked neural responses during emotion interference processing demonstrate specific short-term plasticity driven by type of neurostimulation. C_LIO_LICognitively evoked neural responses are sensitive to variations in mental health symptoms. C_LI

neuroscience↗

bioRxiv2017

Vermont: a multi-perspective visual interactiveplatform for mutational analysis

BackgroundA huge amount of data about genomes and sequence variation is available and continues to grow on a large scale, which makes experimentally characterizing these mutations infeasible regarding disease association and effects on protein structure and function. Therefore, reliable computational approaches are needed to support the understanding of mutations and their impacts. Here, we present VERMONT 2.0, a visual interactive platform that combines sequence and structural parameters with interactive visualizations to make the impact of protein point mutations more understandable.\n\nResultsWe aimed to contribute a novel visual analytics oriented method to analyze and gain insight on the impact of protein point mutations. To assess the ability of VERMONT to do this, we visually examined a set of mutations that were experimentally characterized to determine if VERMONT could identify damaging mutations and why they can be considered so.\n\nConclusionsVERMONT allowed us to understand mutations by interpreting position-specific structural and physicochemical properties. Additionally, we note some specific positions we believe have an impact on protein function/structure in the case of mutation.

bioinformatics↗