Affordable Access

STRUCTURAL FRAME MODEL WITH NUMERICAL REPRESENTATION OF 2D MOLECULAR INFORMATION AND ITS APPLICATION / Acta Phys.-Chim. Sin.

Authors
  • jiaju, zhou
  • zhihong, xu
  • xinjian, yan
  • dunming, sun
  • lingxiao, cao
  • hong, wang
Publication Date
Jan 01, 1993
Source
Institutional Repository of Institute of Process Engineering, CAS (IPE-IR)
Keywords
License
Unknown
External links

Abstract

In the researches of the Quantitative Structure-Activity Relationship, one of the difficult problems is the representation and the extraction of molecule structural informations.In order to solve this problem generally, authors propose a numerical representation method of molecule structural factors-Structural Frame Model method(SFM). First, a model for describing molecule structural factors is defined on a set of compounds being studied. In the set of compounds, there must be a common mole-cular backbone. From the model definitions, the factors reflecting the differences of molecular structures can be completely extracted. Then the whole topological strcuture information can be transfered to a numerical matrix. The activity classification results of 55 sulfonyl urea compounds for variety of target plants (total of 28 systems) give out 24 correct classified systems among the total 28 systems. The successful rate of systems is 26%. For compounds in a system, the correct rate is more than 87%. In order to check the stability and reliability of predictions of SFM method, computer tests on activity predictions are performed, using the leave-N-out approach (N=5). The average correct rate of prediction is 81%. These results show the new method is stable and reliable. / In the researches of the Quantitative Structure-Activity Relationship, one of the difficult problems is the representation and the extraction of molecule structural informations.In order to solve this problem generally, authors propose a numerical representation method of molecule structural factors-Structural Frame Model method(SFM).

Report this publication

Statistics

Seen <100 times