An essential part of the OCHEM platform is the modeling framework. Its m= ain purpose is to provide facilities for the development of predictive comp= utational models for physicochemical and biological properties of compounds= . The framework is integrated with the database of experimental data and in= cludes all the necessary steps required to build a computational model: dat= a preparation, calculation and filtering of molecular descriptors, applicat= ion of machine learning methods and analysis of a models=E2=80=99 performan= ce. This section gives an overview of these features and of the steps requi= red to build a computational model in the OCHEM.
OCHEM modeling framework allows to perform the full cycle of QSAR model = development, which includes:
- Management of datasets with experimental data.
Us= ers can create and manage reusable datasets referred to as baskets= . - Calculation of molecular descriptors.
OCHEM suppo= rts more that 20 types of state-of-the-art molecular descriptors from diffe= rent 3rd party vendors. - Running a machine learning method
- Proper validation protocol of the model
- Calculation of model statistics
- Application of the model to new compounds
- Recalculation of the model based on new experimental e= vidence
Concisely, the main features of the modeling framework within the OCHEM = include:
-
Support of regression and classification models
-
Calculation of various molecular descriptors ranging from molecular fragmen= ts to quantum chemical descriptors. Both whole-molecule and per-atom descri= ptors are supported.
-
Tracking of each compound from the training and validation sets
-
Basic and detailed model statistics and evaluation of model performance on = training and validation sets
-
Assessment of applicability domain of the models and their prediction accur= acy
-
Pre-filtering of descriptors: manual (external) selection, de-correlation f= ilter, Unsupervised Forward Selection (UFS)
-
Various machine learning methods including both linear and non-linear appro= aches
-
N-fold cross-validation and bagging validation of models
-
Multi-learning: models can predict several properties simultaneously
-
Combining data with different conditions of measurements and the data in di= fferent measurement units
-
Distribution of calculations to an internal cluster of Linux and Mac comput= ers
-
Scalability and expendability for new descriptors and machine learning meth= ods
The steps of a typical QSAR research in the OCHEM system and the corresp= onding features are summarized in a diagram in the following figure: