“School of Biological Sciences”

Back to Papers Home
Back to Papers of School of Biological Sciences

Paper   IPM / Biological Sciences / 15336
School of Biological Sciences
  Title:   Identification of molecular features necessary for selective inhibition of B cell lymphoma proteins using machine learning techniques
1.  Ahmad Mani-Varnosfaderani
2.  Marzieh Sadat Neiband
3.  Ali Benvidi
  Status:   Published
  Journal: Molecular diversity
  Year:  2018
  Pages:   1-19
  Supported by:  IPM
Selective inhibition of Bcl-2 and Bcl-xL proteins due to their dual inhibition toxicity plays an important role in treatment of cancer and chemotherapy effectiveness; therefore, in the last decade, discovery of selective inhibitors for Bcl-2 and Bcl-xL proteins has become a significant and important research topic. The present contribution paves the way for characterization of molecular features which induce selectivity for inhibition of Bcl-2 and Bcl-xL. In this line, a total of 1534 molecules related to inhibition of Bcl-2 and Bcl-xL proteins were collected from Binding Database. A diverse set of molecular descriptors was calculated for each molecule, and the best subset of descriptors were selected using variable importance in projection (VIP) approach. The molecules were classified according to their therapeutic targets (Bcl-2/Bcl-xL) and activities. Partial least square-discriminate analysis (PLS-DA) and supervised Kohonen network (SKN) models were utilized to relate the molecular structures of chemicals to their activities and selectivities. According to the VIP-selected descriptors physicochemical properties, such as polarity number, number of branches, size and cyclicity of the molecule, flexibility, functional counts and constitutional descriptors, all affect the activities of Bcl-2 and Bcl-xL inhibitors. The performances of PLS-DA and SKN methods were evaluated based on statistical parameters derived from the confusion matrices. The models were validated using tenfold cross-validation and an external test set. The best statistical results were obtained by implementing the SKN model. The classification rates range from 93.5 to 79.1

Download TeX format
back to top
scroll left or right