Computational protein folding has attracted considerable interest over the years, including molecular simulations and artificial intelligence assisted methods. On the other hand, research and development of quantum computer hardware and software have been thriving recently. In this paper, we report a case study of peptide (PSVKMA) folding based on a two-dimensional lattice model, by using both the blueqat quantum simulator (called AutoQML) and the IonQ quantum device. As a result, it was found that the actual device was still susceptible to noises.
With an increase in an element, the configurational entropy effect stabilizes the multi-element materials. However, the expansion of configuration space and heterogeneous surfaces such as nanoparticles preclude the analytical evaluation of configurational entropy. Then, we implemented the Wang-Landau algorithm, which is one of the Monte-Carlo algorithms, for evaluating the configurational entropy and probing the thermodynamic stable configuration of multi-element materials. The regression equation obtained by density functional theory calculation and multiple regression analysis is used in the energy estimation in the sampling. This method was applied to binary alloys in the bulk and ternary alloy nanoparticles and the obtained features of the stable configuration were discussed.
Prediction of blood-brain barrier permeability for chemicals is one of the key issues in brain drug development. In this study, the effect of using training data relatively similar to the test data was investigated in order to improve the performance of machine learning methods in predicting blood-brain barrier permeability. The results showed that selecting training data with high cosine similarity to the test data improved prediction performance with a smaller number of training data. The best model in this study also showed improved scores on two external test sets to examine generalization performance, outperforming excellent existing models. The cosine similarity method is expected to be effective for predicting the properties of compounds with large diversity and a small number of data.