Obesity and metabolic disorders are rising global health concerns, emphasizing the need for effective dietary interventions. High-viscosity dietary fibers such as bacterial cellulose (BC) and guar gum (GG) have unique properties that may complement each other in modulating gut microbiota and metabolic health. This study investigates their effects in high-fat diet-fed mice. BC and GG increase Bacteroides, which degrade polysaccharides and produce short-chain fatty acids (SCFAs), supporting metabolic health. BC enhances bile acid excretion and enriches Faecalibaculum, Duncaniella, and Paramuribaculum, promoting gut barrier integrity and reducing inflammation, potentially improving bile acid turnover and lipid metabolism. GG more effectively increases butyrate production by enhancing butyrate-producing bacteria, such as Clostridium XIVa and Kineothrix, and promotes Bifidobacterium, strengthening anti-inflammatory effects and gut barrier function. Both fibers upregulate bile acid biosynthesis, but BC's non-fermentable nature leads to higher bile acid excretion, while GG's fermentation causes lower excretion and broader liver metabolic changes. Both fibers reduce body weight, fat accumulation, and cholesterol levels, highlighting their potential in managing obesity and metabolic disorders. The complementary effects of BC and GG underscore the importance of fiber diversity for targeted dietary strategies to improve metabolic health.

2025-01-01·Journal of Molecular Structure

Molecular representation learning based on Transformer with fixed-length padding method

作者: Zhang, Ruimeng ; Wu, Fanhong ; Zou, Yi ; Jin, Meichen ; Wu, Yichu ; Yang, Yang ; Chen, Zijian ; Wang, Zhonghua

Effective mol. representation learning plays an important role in mol. modeling process of drug design, protein engineering, material science and so on.Currently, self-supervised learning models based on the Transformer architecture have shown great promise in mol. representation.However, batch training of Transformer model requires input data of consistent length, while the length of each entrys mol. data (SMILES sequence) is inconsistent, which results in the model being unable to process batches directly.Therefore, corresponding strategies should be proposed to enable the model to smoothly process data with inconsistent length in batches.In this work, we adopt a strategy of head-tail padding and tail padding to obtain fixed-length data, which are employed as inputs for the Transformer encoder and decoder resp., thus overcoming the limitation of the Transformers inability to batch process input data with inconsistent length.In this way, our Transformer-based model can be used for batch training of mol. data, thereby improving the efficiency, accuracy, and simplicity of mol. representation learning.Subsequently, public datasets are used to evaluate the performance of our mol. representation model in predicting mol. property.In the classification and regression tasks, the average ROC-AUC and RMSE values improves by over 10.3% and 3.3% resp. compared to the baseline models.Furthermore, the specific distributions are found after the compressing mol. representation vectors into two-dimensional or three-dimensional space using PCA dimensionality reduction algorithm, instead of random distributions.Our work highlights the potential of Transformer model in batch training for constructing mol. representation model, thus providing new path for AI technol. in mol. modeling.

2023-05-01·Letters in Organic Chemistry

Prediction of Partition Coefficient of Carbamates using GA-MLR and GAANN Methods, and Comparison with Experimental Data

作者: Isfahani, Tahereh Momeni ; Moosavi, Seyedeh Azadeh ; Mohammadinasab, Esmat

Abstract:In the present study, quantum mechanics calculations at the B3LYP theory level and 6- 31G* basis set were carried out to obtain the optimized geometry of carbamates. Then, a comprehensive set of molecular descriptors was computed by using the Dragon software. A genetic algorithm (GA) was also applied to select the suitable variables that resulted in the best-fixed models. The relationship between the molecular descriptors and the partition coefficient of 66 types of carbamates is represented. The molecular descriptors were applied for modeling the multiple linear regression (MLR) and artificial neural network (ANN) methods. The quantitative structure-property relationship models showed that the GA-ANN over the GA-MLR approach resulted in the best outcome. So, the predicted partition coefficient was found to be in good agreement with the experimental partition coefficient. The EEig01x and ALOGP descriptors were applied for modeling the multiple linear regression (MLR) and artificial neural network (ANN) methods. The best model was validated by Q2 LOO, Q2 F1, Q2 F2, Q2 F3, and CCC techniques and external validation parameters for the established theoretical models.

100 项与炔己蚁胺相关的药物交易

登录后查看更多信息