Thus, the analysis of CoMFA and CoMSIA models for Class I+II revealed the points and regions that are highly correlated to the activity of tested compounds.
All of the subsequent models for class I viral fusion proteins are based on the structural data of these two fusion proteins.
Similar to the analysis of idiopathic and LRRK2 PD patients, we used preprocessed datasets to build PLS-DA models for class separation and membership prediction.
To avoid possible statistical artifacts, and to find the analytes (potential biomarkers) which can be used to build PLS-DA models for class discrimination and class membership prediction, we used a preprocessing approach.
After partitioning 211 modeling set compounds into 119 class 1 compounds and 92 class 2 compounds, we developed 843 classification models for class 1 versus class 2, 236 continuous LD50 models for class 1 compounds, and 356 models for class 2 compounds.
The analysis of these three data sets resulted in 252 classification models, as well as 1,207 continuous LD50 models for class 1 compounds and 40 continuous LD50 models for class 2 compounds that satisfied the statistical significance threshold criteria.
However, Nielsen et al pointed out the prediction models for Class II HLAs generally have an AUC 0.10 less than those for Class I HLAs.
35, 62, 63 In IEDB, the binding prediction is made from four PSSM-based models for Class I HLAs, while for Class II HLAs, the result comes from nine models including several PSSM-based models and machine learning models.
Our aim in this study was to develop and optimize a spray dried emulsion containing indomethacin as a model for Class II drugs, Labrasol®/Transuctol® mixture as the oily phase, and maltodextrin as a solid carrier.
The resulting model will be the best model for class prediction.
