[id] => 181
[recorddate] => 2008:08:04:17:00:58
[lastupdate] => 2008:08:04:17:00:58
[type] => conference
[place] => Obernai, France
[subject] => chemistry - general; informatics - applied; mathematics - modeling; mathematics - statistics
[relatedworks] =>
• 4 (some):
  • Statistical assessment of solvent mixtures models used for separation of biological active compounds, ?id=164
  • Model formulation & interpretation - from experiment to theory, ?id=171
  • Structure-property based model for alkanes boiling points, ?id=173
  • The relationship between energy calculations and boiling points of n-alkanes, ?id=231
  • Diagnostic of a QSPR model: aqueous solubility of drug-like compounds, ?id=232
  • Distributing correlation coefficients of linear structure-activity/property models, ?id=242
  • Linear regression modeling and validation strategies for structure-activity relationships, ?id=275
  • Extending characteristic polynomial from graphs to molecules, ?id=303
[file] => ?f=181
[mime] => application/pdf
[size] => 255429
[pubname] => Strasbourg Summer School on Chemoinformatics
[pubinfo] => Louis Pasteur University
[pubkey] => June 22-25
[workinfo] => #3
[year] => 2008
[title] => Statistical approach of structure-activity relationships: a case study
[authors] => Sorana D. BOLBOACĂ, Lorentz JÄNTSCHI
[abstract] =>
An integrated system called Molecular Descriptors Family on Structure-Activity Relationships (MDF-SAR) was developed and used for analysis and quantification of the link between compounds’ structure and measured activity/property in order to obtain structure-activity relationships (SARs). The MDF-SAR approach is able to obtain simple as well as multiple linear regression models between structure (from quantum based descriptors [1] and measured activity/property using a genetic algorithm. The results obtained by using the MDF-SAR approach are online available [2]. A series of methods were proposed for assessment of the obtained models [3]. Starting with the proposed methods some client-server statistical software applications were developed. The statistical approach of structure-activity/property relationships included 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).
References:
[1] Jäntschi L, Bolboacã SD. Int J Quant Chem, 107(8), 1736-1744, 2007.
[2] Jäntschi L, Bolboacã SD, Diudea MV. Int J Mol Sci, 8(11), 1125-1157, 2007.
[3] Bolboacã SD, Jäntschi L, Env Chem Lett, DOI 10.1007/s10311-007-0119-9.
[keywords] => Structure - activity relationships; inferential statistics; models assessment
[acknowledgment] => [MDF] The MDF project was supported through ET36 research project (2005-2007). [MDF-SAR] The MDF-SAR of MDF was supported through ET108 project (2006-2008).