Current Highlight from September 26, 2014
An Empirical Method for Predicting Toxicity?
Scientists from NCTR and the Center for Drug Evaluation and Research developed a novel toxicity modeling approach utilizing 3D NMR-distance based descriptors to identify unique structural alerts (toxicophores) associated with phospholipidosis (PLD). PLD is a common side effect of many drug classes; however, there is a lack of understanding of its molecular mechanism. The 3D-SDAR PLD classification model was based on a training dataset of 328 compounds with publicly available information and was validated on external (proprietary FDA dataset of 294 drugs) and hold-out test sets (20% of training). The classification accuracy for the external test set was 70%. This in silico modeling approach could be utilized to predict which new compounds have a potential to induce PLD; and could be adapted as a general purpose tool to discover structural alerts for other drug side effects and to design new and safer drugs. A manuscript describing the method is available online at Bioorganic & Medicinal Chemistry
.
.For additional information, please contact Rick Beger, Ph.D., Director, Biomarkers and Alternative Models Branch, Division of Systems Biology, FDA/NCTR.
No hay comentarios:
Publicar un comentario