This study presents a transfer learning approach for discovering potential Mg-based superconductors utilizing a comprehensive target dataset.Initially,a large source dataset(Bandgap dataset)comprising approximately∼7...This study presents a transfer learning approach for discovering potential Mg-based superconductors utilizing a comprehensive target dataset.Initially,a large source dataset(Bandgap dataset)comprising approximately∼75k compounds is utilized for pretraining,followed by fine-tuning with a smaller Critical Temperature(T_(c))dataset containing∼300 compounds.Comparatively,there is a significant improvement in the performance of the transfer learning model over the traditional deep learning(DL)model in predicting Tc.Subsequently,the transfer learning model is applied to predict the properties of approximately 150k compounds.Predictions are validated computationally using density functional theory(DFT)calculations based on lattice dynamics-related theory.Moreover,to demonstrate the extended predictive capability of the transfer learning model for new materials,a pool of virtual compounds derived from prototype crystal structures from the Materials Project(MP)database is generated.T_(c) predictions are obtained for∼3600 virtual compounds,which underwent screening for electroneutrality and thermodynamic stability.An Extra Trees-based model is trained to utilize E_(hull)values to obtain thermodynamically stable materials,employing a dataset containing Ehull values for approximately 150k materials for training.Materials with Ehull values exceeding 5 meV/atom were filtered out,resulting in a refined list of potential Mg-based superconductors.This study showcases the effectiveness of transfer learning in predicting superconducting properties and highlights its potential for accelerating the discovery of Mg-based materials in the field of superconductivity.展开更多
Type II diabetes and obesity are two of the most prevalent metabolic disorders effecting a huge population throughout the world. Research over the last decade has unequivocally established considerable molecular links...Type II diabetes and obesity are two of the most prevalent metabolic disorders effecting a huge population throughout the world. Research over the last decade has unequivocally established considerable molecular links between them and hence they are often described in conjugation as ‘diabesity’. The hallmarks of type II diabetes are primarily reduced insulin sensitivity, progressive insulin resistance and consequent hyperinsulinemia.1 Whereas, hyperinsulinemia promotes a plethora of fat synthesis from excess circulating carbohydrate and hyperactive fat storage mechanisms, ultimately inducing hepatic steatosis, myosteatosis and pancreatic steatosis.1 This in turn, aggravates the insulin sensitivity further, inflicting more insulin resistance and even more hyperinsulinemia. Needless to mention that the vicious process in turn, promotes uncontrolled weight gain1 and many secondary metabolic disorders. The prevalent anti-diabetic and anti-obesity medications comes with several limitations ranging from inefficiency to adverse side effects.2 Here, we report an efficient strategy of repositioning previously approved drugs with novel indication in the context of diabesity by investigating deregulated signalling axes affecting patients with both the disorders. Our approach relies extensively on deciphering the strength of gene association in various interactomes, as it is known that within networks, genes linked to similar disease phenotypes tend to be functionally similar3 and remain proximal to each other.4 Moreover, the potential drug targets associated to a disease pathway also cluster proximal to the disease pathways.展开更多
文摘This study presents a transfer learning approach for discovering potential Mg-based superconductors utilizing a comprehensive target dataset.Initially,a large source dataset(Bandgap dataset)comprising approximately∼75k compounds is utilized for pretraining,followed by fine-tuning with a smaller Critical Temperature(T_(c))dataset containing∼300 compounds.Comparatively,there is a significant improvement in the performance of the transfer learning model over the traditional deep learning(DL)model in predicting Tc.Subsequently,the transfer learning model is applied to predict the properties of approximately 150k compounds.Predictions are validated computationally using density functional theory(DFT)calculations based on lattice dynamics-related theory.Moreover,to demonstrate the extended predictive capability of the transfer learning model for new materials,a pool of virtual compounds derived from prototype crystal structures from the Materials Project(MP)database is generated.T_(c) predictions are obtained for∼3600 virtual compounds,which underwent screening for electroneutrality and thermodynamic stability.An Extra Trees-based model is trained to utilize E_(hull)values to obtain thermodynamically stable materials,employing a dataset containing Ehull values for approximately 150k materials for training.Materials with Ehull values exceeding 5 meV/atom were filtered out,resulting in a refined list of potential Mg-based superconductors.This study showcases the effectiveness of transfer learning in predicting superconducting properties and highlights its potential for accelerating the discovery of Mg-based materials in the field of superconductivity.
基金supported by Start-up Research Grant(No.SRG/2019/000221)from Science&Engineering Research Board(SERB),Government of India to D.D.
文摘Type II diabetes and obesity are two of the most prevalent metabolic disorders effecting a huge population throughout the world. Research over the last decade has unequivocally established considerable molecular links between them and hence they are often described in conjugation as ‘diabesity’. The hallmarks of type II diabetes are primarily reduced insulin sensitivity, progressive insulin resistance and consequent hyperinsulinemia.1 Whereas, hyperinsulinemia promotes a plethora of fat synthesis from excess circulating carbohydrate and hyperactive fat storage mechanisms, ultimately inducing hepatic steatosis, myosteatosis and pancreatic steatosis.1 This in turn, aggravates the insulin sensitivity further, inflicting more insulin resistance and even more hyperinsulinemia. Needless to mention that the vicious process in turn, promotes uncontrolled weight gain1 and many secondary metabolic disorders. The prevalent anti-diabetic and anti-obesity medications comes with several limitations ranging from inefficiency to adverse side effects.2 Here, we report an efficient strategy of repositioning previously approved drugs with novel indication in the context of diabesity by investigating deregulated signalling axes affecting patients with both the disorders. Our approach relies extensively on deciphering the strength of gene association in various interactomes, as it is known that within networks, genes linked to similar disease phenotypes tend to be functionally similar3 and remain proximal to each other.4 Moreover, the potential drug targets associated to a disease pathway also cluster proximal to the disease pathways.
