Brazilian Journal of Science

Phytochemical profile and neuropharmacological potential of Erythrina velutina Willd. (Fabaceae): An integrative preclinical review

Fri, 20 Mar 2026 00:00:00 -0300

Medicinal plants constitute an important source of bioactive compounds for the discovery of novel neuroactive agents. Erythrina velutina Willd. (Fabaceae). Popularly known as mulungu, has been traditionally used in the treatment of anxiety, insomnia, and other conditions related to the central nervous system (CNS), which has motivated extensive pharmacological and preclinical investigations. This integrative review critically synthesizes the available evidence on the phytochemical profile and neuropharmacological potential of E. velutina. The literature search was conducted in PubMed, SciELO, Web of Science, and the Virtual Health Library, including preclinical in vitro and in vivo studies, with no language or publication-year restrictions. The analyzed studies indicate a complex phytochemical composition characterized primarily by alkaloids, flavonoids, and other phenolic compounds. Preclinical evidence demonstrates anxiolytic, sedative, anticonvulsant, antioxidant, neuroprotective, and anticholinesterase activities associated with modulation of GABAergic and cholinergic neurotransmission, as well as with reductions in oxidative stress and neuronal protection. The available toxicological data suggest low acute toxicity and the absence of genotoxic effects at the experimentally evaluated doses. However, the lack of standardized extracts, compound-guided pharmacological studies, long-term toxicological assessments, and clinical investigations limits the translational extrapolation of the findings. Therefore, future studies focused on bioprospecting, elucidation of mechanisms of action, and clinical validation are required to consolidate the pharmacological potential of Erythrina velutina.

Analysis of one-carbon compound microbial assimilation pathways and research progress in synthetic biology modification

Baoxin Zhang, Hailei Zhang, Li Zhu, Xiang Weng, Hao Sun — Tue, 31 Mar 2026 00:00:00 -0300

One-carbon (C1) compounds, including methane, methanol, formate, and carbon dioxide, represent promising alternative feedstocks for sustainable biomanufacturing. This comprehensive review systematically analyzes the molecular mechanisms underlying microbial assimilation of C1 compounds, focusing on key metabolic pathways including the ribulose monophosphate (RuMP) pathway, xylulose monophosphate (XuMP) pathway, serine cycle, and the reductive glycine (rGly) pathway. We discuss recent advances in synthetic biology approaches for engineering C1-utilizing microorganisms, including pathway optimization, enzyme engineering, adaptive laboratory evolution, and compartmentalization strategies. Furthermore, we present an analysis of the development of the synthetic biology industry in major Chinese provinces and autonomous regions, including Xinjiang, Gansu, Ningxia, Hunan, and Guangdong. The review highlights the challenges and future directions in developing efficient C1-based cell factories for industrial applications, emphasizing the integration of multi-omics approaches, artificial intelligence, and systems metabolic engineering to enable next-generation C1 biotransformation platforms.

Mathematics-informed machine learning for mapping cell development

Belay Sitotaw Goshu — Tue, 31 Mar 2026 00:00:00 -0300

Single-cell RNA sequencing provides high-resolution snapshots of cellular states, yet descriptive trajectory inference methods are limited to interpolation, struggle with extrapolation, perturbation prediction, and causal mechanism discovery, constraining their utility in predictive developmental biology, cancer therapeutics, and reprogramming. This study introduces and evaluates a Mathematics-Informed Machine Learning (MIML) framework that embeds biological priors mass conservation, attractor dynamics, geometry-aware manifolds, and causal GRN inference into Neural Ordinary Differential Equations to enable continuous, predictive, and mechanistically interpretable modeling of single-cell dynamics. MIML was benchmarked against descriptive baselines (PAGA-like) and standard Neural ODEs across extrapolation, perturbation simulation (knockout/overexpression), out-of-distribution generalization, causal discovery, stage-specific prediction, drug response forecasting, cancer progression reconstruction, and cellular reprogramming efficiency using quantitative metrics (R², MSE, precision/recall/F1, synergy scores, fate probabilities). MIML achieves 1.1-1.3× better prediction/extrapolation, 8-21× gains in perturbation and OOD tasks, ~10-11× improvement in causal GRN accuracy, 82-86% reprogramming success, and 66% HDAC inhibitor response prediction with synergistic combination insights. Overall clinical impact score: 0.82. This is the first framework to jointly enforce biophysical constraints (mass homeostasis, stable attractors), manifold geometry, and directed causal inference within a unified Neural ODE paradigm, yielding unprecedented predictive power and mechanistic insight beyond existing continuous or discrete single-cell models. MIML substantially outperforms existing methods in predictive fidelity, biological plausibility, and translational relevance, establishing a foundation for mechanism-guided single-cell analysis. Prospective interventional validation, spatial/multi-omics integration, stochastic extensions, and interpretability enhancements will further position MIML as a cornerstone for precision biomedicine.