From: Egon W. <ewi...@un...> - 2006-11-13 07:34:39
|
Ok, here's that forward E. ---------- Forwarded Message ---------- Subject: [Cdk-devel] QSAR descriptors: where to put the algorithm implementation... Date: Thursday 26 October 2006 16:11 From: Egon Willighagen <ewi...@un...> To: cdk...@li... Hi all, a short discussion today urged to send around a small email on QSAR descriptors and algorithms that those descriptors use. First, let's define a few concepts. 1. a descriptor is a numerical representation of some chemical property 2. a chemical property is a property inherent to a molecule, bond or atom 3. a chemical property does not have to consist of a numerical vector as a descriptor does The difference between a property and a descriptor is important. Algorithm implementations to calculate a descriptor should be part of the source code of qsar/*/*.java, but the implementations to calculate properties *must* not. The rationale here is that properties *are* interesting outside a QSAR modeling framework too: you want to be able to calculate them without the intention to use the outcome in model building. The latter is what the cdk-qsar module is about. For example, the atomic descriptor 'gasteiger partial charge' should *not* contain the algorithm to calculate this charge. Instead, the QSAR descriptor class source code should call the class which calculates this, and just wrap the outcome. The partial charge is a property. However, another descriptor is might be the first two values of a SVD on the adjacency matrix of the molecule. This algorithm will very like be part of the QSAR descriptor class source code, as the matrix is the property and the svd values are just a numerical representation of it, as QSAR modeling methods generally cannot work with matrices. These SVD values are really descriptors. Now, properties can be, and often are, used as descriptor, which is likely causing the confusion. Hope this clarifies things, and do file bug reports if QSAR descriptor classes have source code internally that calculate properties. Egon -- CUBIC blog: http://chem-bla-ics.blogspot.com/ ------------------------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ Cdk-devel mailing list Cdk...@li... https://lists.sourceforge.net/lists/listinfo/cdk-devel ------------------------------------------------------- -- CUBIC blog: http://chem-bla-ics.blogspot.com/ |