RESUMO
The removal of carbonyl sulfide (COS) commonly contained in natural gas is of great significance but still very challenging via a widely employed absorption process due to its low reactivity and solubility in various commercial solvents. Artificial intelligence (AI) is playing an increasingly important role in the exploration of desulfurization solvents. However, practically feasible AI models still lack a thorough understanding of the reaction mechanisms. Machine learning (ML) models established on chemical mechanisms exhibit enhanced chemical interpretability and prediction performance. In this study, we constructed a series of solvent molecules with varying functional groups, including linear aliphatic amines, cyclic aliphatic amines, and aromatic amines and proposed a three-step reaction pathway to dissect the effects of charge and steric hindrance of different substituents on their reaction rates with COS. Chemical descriptors, based on electrostatic potential (ESP), average local ionization energy (ALIE) theory, Hirshfeld charges, and Fukui functions, were used to correlate and predict the electrophilic reactivity of amine groups with COS. Substituents influence the reaction rate by changing the attraction interaction of amine groups to COS molecules and the electron rearrangement in the electrophilic reaction. Furthermore, they have more pronounced steric effects on the reaction rate in the linear amines. The descriptors N_ALIE and q(N) were found to be crucial in predicting the reactivity of amine groups with COS. Present study provides a comprehensive understanding of the reaction mechanisms of COS with amine compounds, offers specific chemical principles for the development of chemistry-driven ML models, sheds light on other types of electrophilic reactions occurring on amine and phosphine groups, and guides the development of chemical solvents in gas absorption processes.
RESUMO
ABSTRACT: Tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) and TRAIL-receptor-2 (TRAIL-R2) are associated with atherosclerosis. This meta-analysis aimed to investigate the potential association between TRAIL/TRAIL-R2 with mortality or cardiovascular (CV) events. PubMed, Embase, and Cochrane Library were searched for reports published up to May 2021. Reports were included when the association between TRAIL or TRAIL-R2 and mortality or CV events was reported. Considering the heterogeneity between studies, we used the random-effects model for all analyses. Ultimately, the meta-analysis included 18 studies (16,295 patients). The average follow-up ranged from 0.25 to 10 years. Decreased TRAIL levels were negatively associated with all-cause mortality [rank variable, hazard ratio (HR), 95% CI, 2.93, 1.94-4.42; I2 = 0.0%, Pheterogeneity = 0.835]. Increased TRAIL-R2 levels were positively associated with all-cause mortality (continuous variable, HR, 95% CI, 1.43, 1.23-1.65; I2 = 0.0%, Pheterogeneity = 0.548; rank variable, HR, 95% CI, 7.08, 2.70-18.56; I2 = 46.5%, Pheterogeneity = 0.154), CV mortality (continuous variable, HR, 95% CI, 1.33, 1.14-1.57; I2 = 0.0%, Pheterogeneity = 0.435), myocardial infarction (continuous variable, HR, 95% CI, 1.23, 1.02-1.49; rank variable, HR, 95% CI, 1.49, 1.26-1.76; I2 = 0.7%, Pheterogeneity = 0.402), and new-onset heart failure (rank variable, HR, 95% CI, 3.23, 1.32-7.87; I2 = 83.0%, Pheterogeneity = 0.003). In conclusion, decreased TRAIL was negatively associated with all-cause mortality, and increased TRAIL-R2 was positively associated with all-cause mortality, CV mortality, myocardial infarction, and heart failure.