Confidence Assessment of Candidate Drug Property Predictions by Subspace Mapping Methods

AXEL JUAN SOTO; GUSTAVO ESTEBAN VAZQUEZ; MARC STRICKERT; IGNACIO PONZONI

Santiago de Chile

Conferencia; ISCB - Latin America 2012; 2012

International Society for Computational Biology (ISCB)

We present a method for assisting virtual screening of drugs during the early stages of the drug development process. This methodology is proposed to improve the reliability of QSPR prediction. First, a transformation is sought for mapping a high-dimensional space defined by potentially redundant or irrelevant molecular descriptors into a low-dimensional target-related space. Second, we apply an applicability domain model on the low-dimensional space for assessing confidentiality of compound classification. By a probabilistic framework our approach identifies poorly represented compounds in the training set and space regions where the uncertainty about the correct class is higher than normal.