Biology subjects

de Oliveira, R.

Publications and source records attributed to de Oliveira, R..

3 recordsLinked to original sources

bioRxiv2025

CMG2 interaction with actin is required for growth factor-induced chemotaxis in endothelial cells.

Capillary morphogenesis gene 2 (CMG2/ANTXR2) is a cell surface receptor that contributes to corneal angiogenesis and is essential for orienting endothelial cell chemotaxis in growth factor gradients. However, the mechanism by which CMG2 transmits signals to orient endothelial cells is unknown. Here, we use affinity proteomics to identify proteins that interact with CMG2. This led us to investigate CMG2 interaction with F-actin in the context of serum and individual growth factors (bFGF/PDGF/VEGF) in both primary and immortalized endothelial cells. We also measured different matrix-protein-mediated changes in chemotaxis. We find that CMG2 function is matrix dependent, and that CMG2-actin interaction is required for serum- and growth-factor-induced chemotaxis. Inhibiting chemotaxis with CMG2 antagonists (collagen 6, penta-galloyl-glucose, or anthrax toxin protective antigen; PA) leads to the release of actin cytoskeleton from CMG2. Furthermore, CMG2 co-localizes with actin at the leading edge during live migration. This colocalization is disrupted by PA, which also disrupts directional movement. Finally, we demonstrate that the conserved actin binding domain of CMG2 directly binds to F-actin in vitro where it bundles F-actin, suggesting a mechanism by which it enables directional migration.

molecular biology↗

bioRxiv2025

A Study of Calibration as a Measurement of Trustworthiness of Large Language Models in Biomedical Research

ObjectivesTo assess the calibration of 9 large language models (LLMs) within biomedical natural language processing (BioNLP) tasks, furthering understanding of trustworthiness and reliability in real-world settings. Materials and MethodsFor each LLM, we collected responses and corresponding confidence scores for all 13 datasets (grouped into 6 tasks) of the Biomedical Language Understanding & Reasoning Benchmark (BLURB). Confidence scores were assigned using 3 strategies: Verbal, Self-consistency, Hybrid. For evaluation, we introduced Flex-ECE (Flexible Expected Calibration Error): a novel adaptation of ECE that accounts for partial correctness in model responses, allowing for a more realistic assessment of calibration in language-based settings. Two post-hoc calibration techniques--isotonic regression and histogram binning--were evaluated. ResultsAcross tasks, mean calibration ranged from 23.9% (Population-Intervention-Comparison-Outcome extraction) to 46.6% (Relation Extraction). Across LLMs, Medicine-Llama3-8B had the best mean overall calibration (29.8%); Flan-T5-XXL had the highest ranking on 5/13 datasets. Across strategies, self-consistency (mean: 27.3%) had better calibration than Verbal (mean: 42.0%) and Hybrid (mean: 44.2%). Post-hoc methods substantially improved calibration, with best mean calibrated Flex-ECEs ranging from 0.1% to 4.1%. DiscussionThe poor out-of-the-box calibration of LLMs poses a risk to trustworthy deployment of such models in real-world BioNLP applications. Calibration can be improved post-hoc and is a recommended practice. Non-binary metrics for LLM evaluation such as Flex-ECE provide a more realistic assessment of trustworthiness of LLMs, and indeed any model that can be partially right/wrong. ConclusionThis study shows that out-of-the-box calibration of LLMs is very poor, but traditional post-hoc calibration techniques are useful to calibrate LLMs.

bioinformatics↗

bioRxiv2024

A probabilistic view of forbidden links: their prevalence and their consequences for the robustness of mutualistic networks

The presence in ecological communities of unfeasible species interactions, termed forbidden links, due to physiological or morphological exploitation barriers has been long debated, but little direct evidence has been found. Forbidden links are likely to make ecological communities less robust to species extinctions, stressing the need to assess their prevalence. Here, we used a dataset of plant-hummingbird interactions, coupled with a Bayesian hierarchical model, to assess the importance of exploitation barriers in determining species interactions. We found evidence for exploitation barriers between flowers and hummingbirds across the 32 studied communities, however, the proportion of forbidden links changed drastically among communities, because of changes in trait distributions. The higher the proportion of forbidden links, the more they decreased network robustness, because of constraints on interaction rewiring. Our results suggest that exploitation barriers are not rare in plant-hummingbird communities and have the potential to limit the rescue of species experiencing partner extinction.

ecology↗