Entradas recentes
AI2SM: Artificial Intelligence to expose plant secondary metabolism
( 2026 ) Ribeiro, João Manuel Capela Araújo; Dias, Oscar; Rocha, Miguel; Ridder, Dick de
Estima-se que as plantas produzam mais de um milhão de metabolitos, mas apenas cerca de 0,1% das suas vias biossintéticas foram elucidadas. Esta lacuna é particularmente acentuada no metabolismo especializado (secundário), apesar do valor destes metabolitos como produtos farmacêuticos, agroquímicos, nutracêuticos e cosméticos. A caracterização das vias requer a reconstrução dos passos de síntese química e a atribuição de cada transformação à reação enzimática subjacente. A retrossíntese automatizada permite prever passos químicos até um metabolito-alvo sem evidência experimental completa, enquanto as ferramentas de anotação de proteínas identificam enzimas e inferem os tipos de reação que catalisam através da previsão do Enzyme Commission number (EC). Embora existam várias ferramentas de retrossíntese e anotação enzimática baseadas em inteligência artificial (IA), regras ou similaridade, estas raramente foram comparadas de forma sistemática ou integradas num quadro unificado. Para colmatar esta lacuna, foi desenvolvido o DeePlants, uma plataforma modular que usa IA para apoiar a elucidação do metabolismo especializado das plantas, integrando anotação funcional de enzimas, associação metabolito-via e previsão retrossintética num pacote unificado. Este trabalho contribui para o estudo de representações computacionais de proteínas e produtos naturais, avaliando a sua eficácia na previsão de números EC, especificidade enzima-substrato e associação a vias biossintéticas, estabelecendo uma base sólida para a previsão da função enzimática e a elucidação de vias. Nos nossos benchmarks, os modelos de linguagem de proteínas superaram codificações one-hot e BLASTp na previsão de números EC, sobretudo na ausência de homólogos. Para associações metabolitovia, a Morgan fingerprint com contagens foi a representação mais eficaz. Para a previsão de especificidade enzima-substrato, uma nova abordagem que combina este fingerprint, embeddings de modelos de linguagem de proteínas e gradient boosting superou os melhores métodos de referência. Adicionalmente, com o DeepMol, uma plataforma de aprendizagem máquina automatizada para química computacional, identificámos combinações ideais de modelos e representações de compostos, com o classificador Ridge a superar o estado da arte e a prever precursores biossintéticos para mais de 75% dos metabolitos vegetais na LOTUS-DB.
Engineering melanin nanoplatforms: from advanced biomedical applications to synergistic cancer therapy
( 2026 ) Caldas, Mariana Elisabete Rodrigues; Correlo, V. M.; Reis, R. L.
Cancer remains one of the leading causes of death worldwide and continues to represent a major global health challenge, demanding the development of more effective, targeted, and less toxic therapeutic strategies. Among malignant bone tumors, osteosarcoma is particularly aggressive and is often associated with limited treatment options, systemic toxicity, and poor outcomes in advanced or recurrent cases. These limitations highlight the urgent need for innovative therapeutic approaches capable of improving treatment efficacy while reducing systemic side effects. In this context, cancer theranostics and localized treatment strategies have emerged as promising avenues, enabling simultaneous diagnosis, targeted drug delivery, and externally triggered therapeutic activation. Melanin is a naturally occurring biopolymer acknowledged by outstanding biocompatibility, biodegradability, photothermal and optoelectronic properties, making it a promising material for biomedical applications. Thus, this thesis explores the potential of natural melanin nanoparticles (MNPs) as sustainable and multifunctional platforms for cancer theranostics and localized therapies, with a particular focus on osteosarcoma treatment. The work begins with a comprehensive review of melanin structure, physicochemical properties, and recent advances in the biomedical applications of melanin and melanin-like nanoparticles, including imaging, drug delivery, and light-triggered therapies. Natural MNPs extracted from Sepia officinalis ink sacs were then developed and functionalized to enable biointerface engineering and targeted biomedical applications, demonstrating high cytocompatibility, hemocompatibility, and versatility for bioreceptor conjugation. The therapeutic potential of MNPs was further investigated in light-activated cancer therapies. MNPs were employed as intracellular nanocarriers for a photosensitizer (rhodamine B isothiocyanate) in photodynamic therapy, enabling enhanced cellular uptake and efficient photoinduced cytotoxicity under both normoxic and hypoxic conditions. In parallel, MNPs were explored as dual-function agents for combined chemo-photothermal therapy, serving simultaneously as doxorubicin nanocarriers and efficient photothermal transducers under near-infrared irradiation, resulting in synergistic anticancer effects in osteosarcoma models. To enable localized and controlled therapeutic delivery, injectable gellan gum–based hydrogel systems incorporating MNPs and doxorubicin were developed and evaluated in physiologically relevant three-dimensional osteosarcoma models. These hybrid systems exhibited tunable mechanical properties, controlled drug release, and effective photothermal performance, leading to substantial tumor cell reduction. The therapeutic efficacy and underlying mechanisms were further validated in a vascularized chorioallantoic membrane model, confirming the potential of hydrogel-mediated chemo-photothermal therapy in complex biological environments. Overall, this thesis demonstrates that natural melanin nanoparticles, alone or integrated into hydrogel-based systems, constitute sustainable, biocompatible, and multifunctional platforms for imaging-guided and light-triggered cancer therapies. These findings support the translational potential of melanin-based materials as versatile tools for localized cancer treatment and future theranostic applications.
Drug repurposing of dopaminergic drugs to inhibit ataxin-3 aggregation
( 2023 ) Figueiredo, Francisco; Sárkány, Zsuzsa; Silva, Alexandra; Campos, Daniela Vilas Boas da Silva; Maciel, P.; Castro, Andreia Cristiana Teixeira; Martins, Pedro M.; Macedo-Ribeiro, Sandra
The accumulation of mutant ataxin-3 (Atx3) in neuronal nuclear inclusions is a pathological hallmark of Machado-Joseph disease (MJD), also known as Spinocerebellar Ataxia Type 3. Decreasing the protein aggregation burden is a possible disease-modifying strategy to tackle MJD and other neurodegenerative disorders for which only symptomatic treatments are currently available. We performed a drug repurposing screening to identify inhibitors of Atx3 aggregation with known toxicological and pharmacokinetic profiles. Interestingly, dopamine hydrochloride and other catecholamines are among the most potent inhibitors of Atx3 aggregation in vitro. Our results indicate that low micromolar concentrations of dopamine markedly delay the formation of mature amyloid fibrils of mutant Atx3 through the inhibition of the earlier oligomerization steps. Although dopamine itself does not cross the blood-brain barrier, dopamine levels in the brain can be increased by low doses of dopamine precursors and dopamine agonists commonly used to treat Parkinsonian symptoms. In agreement, treatment with levodopa ameliorated motor symptoms in a C. elegans model of MJD. These findings suggest a possible application of dopaminergic drugs to halt or reduce Atx3 accumulation in the brains of MJD patients.