Collective recognition strategy for estimating hepatoprotector activity of various chemical compounds in increasing liver repair potential
Autor: | L. A. Radkevich, A. S. Kabankin |
---|---|
Rok vydání: | 2013 |
Předmět: | |
Zdroj: | Pharmaceutical Chemistry Journal. 47:181-186 |
ISSN: | 1573-9031 0091-150X |
DOI: | 10.1007/s11094-013-0922-5 |
Popis: | The relationship between molecular characteristics and hepatoprotector activity (increasing liver repair potential) was investigated for a diverse series of chemical compounds. Application of the nearest-neighbor method using geometric descriptors to a training set including 44 compounds (23 active and 21 inactive) gave combinations of the descriptors that classified correctly 79 – 91% of the training-set compounds. It was shown that a classifier that classified correctly all training-set compounds could be constructed by applying a simple voting procedure to results obtained for various combinations of descriptors. The effectiveness of the proposed approach for predicting the activity of untested compounds was proved by cross-validation with groups of compounds removed from the training set. |
Databáze: | OpenAIRE |
Externí odkaz: |