Lorentz JÄNTSCHI (lori) works ?id=103
- [id] => 103
- [recorddate] => 2007:01:14:17:16:16
- [lastupdate] => 2007:01:14:17:16:16
- [type] => conference
- [place] => Sofia, Bulgaria
- [subject] => chemistry - biochemistry; chemistry - computational; engineering - agriculture; informatics - models implementation; mathematics - modeling
- [relatedworks] =>
- 2 (average):
- Molecular descriptors family on structure-activity relationships: modeling herbicidal activity of substituted triazines class, ?id=87
- [file] => ?f=103
- [mime] => application/pdf
- [size] => 344726
- [pubname] => International Symposium on Organic Chemistry, December 9-12, 2006
- [pubinfo] => Avangard Prisma
- [pubkey] => ISBN 954-323-243-1 & 978-954-323-243-7
- [workinfo] => p. 48-49 [Plenary oral presentation], December 9-12
- [year] => 2006
- [title] => Activity characterization of triazines analogues: statistical parameters for models assessment
- [authors] => Sorana D. BOLBOACĂ, Lorentz JÄNTSCHI
- [abstract] =>
Correlation coefficients and associated squared values are used as parameters in validation of structure-activity relationship models. By using of the molecular descriptors family on structure-activity relationship method  the herbicidal activity of a sample of triazines analogues was modelled . A number of three multivariate models proved to have estimation and prediction abilities .
Starting from the hypothesis that the measured activity of triazines analogue is a semi-quantitative variable, the aim of the research was to analyzed the three previously reported models by using the Pearson, Spearman, Kendall’s and Gamma correlation coefficients.
The structure-activity relationship models were previously reported . The measured herbicidal activity of triazines analogues and the value estimated by the previously reported models were investigated by using the Pearson, Spearman, Kendall's τa, τb, τc and Gamma correlation coefficients (rPrs, ρSpm, τKen,a, τKen,b, τKen,c, Γ) and associated squared correlation coefficient (rPrs2, ρSpm2, τKen,a2, τKen,b2, τKen,c2, Γ2).
The results of investigation, express as correlation coefficients and associated 95% confidence intervals, squared correlation coefficient and Student’s t, respectively the parameter of the Z test were calculated and analyzed.
The correlation coefficients obtained with all methods were statistical significant (p < 0.0001). The correlation coefficients vary according with the model as follow:
• Model with two descriptors: from r = 0.6889 (τKen,c) - 95%CI [0.4370-0.8405] to r = 0.9855 (rPrs) - 95% CI [0.9694-0.9931];
• Model with three descriptors: from 0.7489 (τKen,c) - 95%CI [0.5321-0.8734] to the value equal with 0.9883 (rPrs) - 95% CI [0.9752-0.9944];
• Model with four descriptors: from 0.7444 (τKen,c) - 95%CI [0.5248-0.8710] to the value equal with 0.9925 (rPrs) - 95% CI [0.9841-0.9964].
If there is considered that the herbicidal activity of triazines analogues is a quantitative variable, the Pearson correlation coefficients is the statistical parameter that must be used in evaluation of relationship between the measured and estimated activity.
Considering the measured activity of triazines analogues as a semi-quantitative variable, a rank correlation coefficient is the statistical parameter able to provide more reliable estimation rather than Pearson correlation coefficient. However, which is the proper rank correlation coefficients that can be use in this circumstance? The comparisons between the rank correlation coefficients are discussed and the further plan of research is highlighted.
 Jäntschi L. Molecular Descriptors Family on Structure Activity Relationships 1. Review of the Methodology. Leonardo Electronic Journal of Practices and Technologies 2005;6:76-98.
 Bolboacă S., Jäntschi L. Molecular Descriptors Family on Structure-Activity Relationships: Modeling Herbicidal Activity of Substituted Triazines Class. Bulletin of University of Agricultural Sciences and Veterinary Medicine - Agriculture 2006;62:35-40.
- [keywords] => Triazines Analogues; Statistical Models; Correlation Coefficients
- [acknowledgment] => The research was partly supported by UEFISCSU Romania through project ET108/2006.