Pedro Giesteira Cotovio

Most machine learning methods that explore protein-protein interactions rely exclusively on the network of protein interactions as input data. However, protein repositories are rich in information describing multiple protein aspects, and ignoring this wealth of information may be detrimental to predictive performance. In this paper, we investigate how injecting domain knowledge – the molecular functions proteins perform, the biological processes they are involved in and their cellular location – impacts the performance of graph neural networks on two fundamental predictive tasks: protein-protein interaction prediction and protein function prediction. We propose a novel approach that injects protein-level information in the form of knowledge graph embeddings computed over the Gene Ontology into state-of-the-art graph neural networks, replacing ad hoc strategies for node features. We assessed the impact of our knowledge injection approach in two popular benchmarks: the Open Graph Benchmark (based on STRING) and the Human Reference Interactome. Experiments conducted with ten state-of-the-art graph neural network approaches and five representative knowledge graph embedding methods demonstrated that domain knowledge injection significantly improves performance on graph neural networks in both protein function and protein-protein interaction prediction.

Knowledge Graphs (KG) are the backbone of many data-intensive applications since they can represent data coupled with its meaning and context. Aligning KGs across different domains and providers is necessary to afford a fuller and integrated representation. A severe limitation of current KG alignment (KGA) algorithms is that they fail to articulate logical thinking and reasoning with lexical, structural, and semantic data learning. Deep learning models are increasingly popular for KGA inspired by their good performance in other tasks, but they suffer from limitations in explainability, reasoning, and data efficiency. Hybrid neurosymbolic learning models hold the promise of integrating logical and data perspectives to produce high-quality alignments that are explainable and support validation through human-centric approaches. This paper examines the current state of the art in KGA and explores the potential for neurosymbolic integration, highlighting promising research directions for combining these fields.

The spread of pathogenic and food spoilage microorganisms through the food chain still faces major mitigation challenges, despite modern advances. Although multiple cleaning and disinfection procedures are available for microbial load reduction in food-related settings, microbes can still remain on surfaces, equipment, or machinery, especially if they have the ability to form biofilms. The present study assessed the biofilm-forming properties of pure and mixed cultures of foodborne and spoilage bacteria (Listeria monocytogenes, Enterococcus faecalis, Aeromonas hydrophila, Brochothrix thermosphacta), using polystyrene and stainless steel contact surfaces. Subsequently, the antimicrobial and antibiofilm properties of Thymus vulgaris and Cymbopogon flexuosus essential oils—EOs—were evaluated against these bacteria. Moreover, in silico prediction of the absorption and toxicity values of the EOs’ major constituents was also performed, perceiving the putative application in food-related settings. Overall, biofilm formation was observed for all microbes under study, at different temperatures and both contact surfaces. In polystyrene, at 25 °C, when comparing pure with mixed cultures, the combination Listeria–Aeromonas achieved the highest biofilm biomass. Moreover, at 4 °C, increased biofilm formation was detected in stainless steel. Regarding thyme, this EO showed promising antimicrobial features (especially against A. hydrophila, with a MIC of 0.60 µg/µL) and antibiofilm abilities (MBEC of 110.79 µg/µL against L. monocytogenes, a major concern in food settings). As for lemongrass EO, the highest antimicrobial activity, with a MIC of 0.49 µg/µL, was also observed against L. monocytogenes. Overall, despite promising results, the in situ effectiveness of these essential oils, alone or in combination with other antimicrobial compounds, should be further explored.

Aromatase inhibitors (AI) are extensively used as adjuvant endocrine therapy in post-menopausal women with hormone receptor-positive early breast cancer (HR+ EBC), but their impact on bone health is not negligible. This work aimed to assess bone loss, fracture incidence, and risk factors associated with these events, as well as the prognostic influence of fractures. We have conducted a retrospective cohort study of women with HR+ EBC under adjuvant therapy with AI, during a 3-year period. Four-hundred-and-fifty-one eligible women were reviewed (median age 68 years). Median time under AI was 40 months. A fracture event occurred in 8.4%, mostly in the radium and femoral neck and in older women (mean 74 vs. 68 years, p = 0.006). Age (OR 1.01, 95% CI 1.01–1.07, p = 0.024) and time under AI (OR 1.02, 95% CI 1.00–1.04, p = 0.037) were independent predictors of fracture, with a fair discrimination (AUC 0.71). Analysis of disease-free survival according to fracture event varied between groups, disfavoring the fracture cohort (at 73 months, survival 78.6%, 95% CI, 47.6–92.4 vs. 95.6%, 95% CI, 91.2–97.8, p = 0.027). The multivariate model confirmed the prognostic impact of fracture occurrence (adjusted HR of 3.17, 95% CI 1.10–9.11; p = 0.032). Bone health is often forgotten, despite its great impact in survivorship. Our results validate the pathophysiologic link between EBC and bone metabolism, which translates into EBC recurrence. Further research in this area may help refine these findings. Moreover, early identification of women at higher risk for fractures is warranted.