|
|
|
|
CONTRIBUTORS:
|
|
|
CONFERENCE NAME:
|
|
|
CONF. LOCATION:
|
Plovdiv, Bulgaria
|
|
CONFERENCE YEAR:
|
2008
|
|
PUB TYPE:
|
Conference Presentation
|
|
SUBJECT(S):
|
chemistry - computational; mathematics - modeling; mathematics - number theory; mathematics - statistics
|
|
DISCIPLINE:
|
Mathematics
|
|
HTTP:
|
http://lori.academicdirect.org/works/?f=199
|
|
LANGUAGE:
|
English
|
|
PUB ID:
|
103-445-899
(Last edited on
2008/10/17 02:53:29 GMT-6)
|
|
SPONSOR(S):
|
|
|
ABSTRACT:
General validation principles for quantitative structure-activity relationship (qSAR) models in the context of chemical regulation were developed due to the importance and implication in of these methods in drug design. A brief analysis of different techniques used in validation of multiple linear regression models is reviewed. The hierarchical steps in models validation are highlighted and a validation technique is proposed. The following statistical approaches are considered: correlation analysis (Pearson, Spearman, Kendall and Gamma coefficients as parameters and associated significance levels), regression analysis (leave-one-out cross-validation and determination coefficients), and other inferential statistics (cross correlation coefficients, training vs. test experiment, correlated correlations analysis). The proposed statistical validation technique is exemplified on a qSAR model obtained by applying the molecular descriptors family on the structure-activity relationship approach.
|
|
|
|
STATISTICS
|
|
Click on # to view
|
|
Citations
|
|
2
|
|
References
|
|
4
|
|
Comments
|
|
0
|
|
Quality
|
|
0/0.00
|
|
Interest
|
|
0/0.00
|
|
View(er)s
|
|
1/253
|
|
|
|
|
|
|
| Prev |
Next |
|
