Multi-modal pre-training models have been intensively explored to bridge vision and language in recent years. However, most of them explicitly model the cross-modal interaction between image-text pairs, by assuming that there exists strong semantic correlation between the text and image modalities. Since this strong assumption is often invalid in real-world scenarios, we choose to implicitly model the cross-modal correlation for large-scale multi-modal pretraining, which is the focus of the Chinese project 'Wen-Lan' led by our team. Specifically, with the weak correlation assumption over image-text pairs, we propose a twotower pre-training model called BriVL within the crossmodal contrastive learning framework. Unlike OpenAI CLIP that adopts a simple contrastive learning method, we devise a more advanced algorithm by adapting the latest method MoCo into the cross-modal scenario. By building a large queue-based dictionary, our BriVL can incorporate more negative samples in limited GPU resources. We further construct a large Chinese multi-source imagetext dataset called RUC-CAS-WenLan for pre-training our BriVL model. Extensive experiments demonstrate that the pre-trained BriVL model outperforms both UNITER and OpenAI CLIP on various downstream tasks.
Five different proteases were used to hydrolyze the swim bladders of Nibea japonica and the hydrolysate treated by neutrase (collagen peptide named SNNHs) showed the highest DPPH radical scavenging activity. The extraction process of SNNHs was optimized by response surface methodology, and the optimal conditions were as follows: a temperature of 47.2 °C, a pH of 7.3 and an enzyme concentration of 1100 U/g, which resulted in the maximum DPPH clearance rate of 95.44%. Peptides with a Mw of less than 1 kDa (SNNH-1) were obtained by ultrafiltration, and exhibited good scavenging activity for hydroxyl radicals, ABTS radicals and superoxide anion radicals. Furthermore, SNNH-1 significantly promoted the proliferation of HUVECs, and the protective effect of SNNH-1 against oxidative damage of H2O2-induced HUVECs was investigated. The results indicated that all groups receiving SNNH-1 pretreatment showed an increase in GSH-Px, SOD, and CAT activities compared with the model group. In addition, SNNH-1 pretreatment reduced the levels of ROS and MDA in HUVECs with H2O2-induced oxidative damage. These results indicate that collagen peptides from swim bladders of Nibea japonica can significantly reduce the oxidative stress damage caused by H2O2 in HUVECs and provides a basis for the application of collagen peptides in the food industry, pharmaceuticals, and cosmetics.
Fucoxanthin, a xanthophyll carotenoid abundant in brown algae, is reported to have several biological functions, such as antioxidant, anti-inflammatory, and anti-tumor activities, in mice. We investigated the effects and mechanisms of fucoxanthin in the mixture oleate/palmitate = 2/1(FFA)-induced nonalcoholic fatty liver disease (NAFLD) cell model in this study. The results showed that the content of superoxide dismutase in the FFA group was 9.8 ± 1.0 U/mgprot, while that in the fucoxanthin high-dose (H-Fx) group (2 μg/mL) increased to 22.9 ± 0.6 U/mgprot. The content of interleukin-1β in the FFA group was 89.3 ± 3.6 ng/mL, while that in the H-Fx group was reduced to 53.8 ± 2.8 ng/mL. The above results indicate that fucoxanthin could alleviate the FFA-induced oxidative stress and inflammatory levels in the liver cells. Oil red-O staining revealed visible protrusions and a significant decrease in the number of lipid droplets in the cytoplasm of cells in the fucoxanthin group. These findings on the mechanisms of action suggest that fucoxanthin can repair FFA-induced NAFLD via the adenosine monophosphate-activated protein kinase (AMPK) signaling pathway and nuclear factor erythroid-2-related factor 2-mediated (Nrf2) signaling pathway, as well as by downregulating the expression of the Toll-like receptor 4-mediated (TLR4) signaling pathway. Fucoxanthin exhibited alleviating effects in the FFA-induced NAFLD model and could be explored as a potential anti-NAFLD substance.
Background: Marine fish meat has been widely used for the extraction of bioactive peptides. This study was aimed to optimize the preparation of monkfish muscle peptides (LPs) using response surface methodology (RSM) and explore the antioxidant activities of <1 kDa LPs. Methods: Peptides were prepared from the muscles of monkfish (Lophius litulon), and five proteases were tested to hydrolyze muscle proteins. The hydrolysate that was treated using neutrase showed the highest degree of hydrolysis (DH) and 1,1-diphenyl-2-picrylhydrazyl (DPPH) scavenging activities. Results: The optimized conditions were as follows: water/material ratio of 5.4:1, a time span of 5 h, pH of 7.0, enzyme concentration of 2000 U/g, and temperature of 45 °C; the maximum DPPH scavenging activity and DH were 92.861% and 19.302%, respectively. LPs exhibited appreciable antioxidant activities, including DPPH radical, hydroxyl radical, 2,2′-azinobis-3-ethylbenzthiazoline-6-sulphonate (ABTS) radical, and superoxide anion scavenging activities. LPs attenuated H2O2-related oxidative injury in RAW264.7 cells, reduced the reactive oxygen species (ROS) and malondialdehyde (MDA) levels, and increased the superoxide dismutase (SOD), glutathione peroxidase (GSH-Px), and catalase (CAT) levels. Conclusion: We concluded that LPs could be an ideal source of bioactive peptides from monkfish and also have pharmaceutical potential.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.