TopKat system

Prival, M.. Evaluation of the TOPKAT system for predicting the carcinogenicity of chemicals. Environ. Mol. Mutagen. 2001, 37, 55-69. 

The example of the toxicity estimation using QSAR is based on the TOPKAT system developed by Health Designs, Inc., and is the computer program... 

A number of commercial expert systems have been applied to screen drug libraries. For instance, DEREK, TOPKAT, MultiCASE, and many other systems all have possibilities in this regard. However, it should be noted that for broad screening only compounds with toxicity associated with them can be identified, and hence these are very crude measures of hazard assessment. The use of expert systems to screen libraries is fraught with dangers, not least that no performance statistics are available for these systems being used for such an application. It is also highly probable that the vast majority of predic-... 

There are two commercially available expert systems that rely on a statistical approach TOPKAT, which was developed by Health Designs Inc. but is now owned by Accelrys [18], and CASE/MultiCASE, developed at Case Western University [31]. 

There are a growing number of tools to assess applicability domain, and a number of expert systems, for example, TOPKAT and MultiCASE, have their own measures of fit. These need to be developed and their application to larger drug libraries demonstrated. 

TOPKAT Automated Rule-Induction Systems Accelrys Ltd. www.accelrys.com/products/topkat/ mutagenicity, teratogenicity, membrane irritation, neurotoxicity, and other effects Carcinogenicity,... 

ARI systems make quantitative predictions, for example, by providing a probability value of carcinogenicity being induced by a molecule or a quantitative prediction of an acute toxicity. Two widely used ARI systems are Toxicity Prediction by Komputer-Assisted Technology (TOPKAT) (Enslein, 1988 Enslein et al 1994) and Multiple Computer Automated Structure Evaluation... 

The expert system TOPKAT has been evaluated in several studies (Snyder et al. 2004 Dearden 2003 Patlewicz et al. 2003 Cariello et al. 2002 Durham and Pearl 2001 Richard 1998 Benfenati and Gini 1997 Enslein et al. 1994). In these evaluations all statistical parameters like sensitivity, specificity, positive and negative predictivity and the overall concordance range from 40 % to 73 % for the various endpoints. 

Unfortunately no modification of the training set or the prediction models is possible for the QSAR-based expert system TOPKAT. 

It is conceivable that quantitative structure-activity (QSAR) approaches (e.g., TOPKAT see Chapter 7) could be applied to predict response levels for uncharacterized contaminants for use in the HI approach. Further, specific submodels existing (e.g., that for developmental toxicity) could be applied to estimate system-specific response levels for application in the IT D approach. To our knowledge, there are no computer-assisted programs available that can automate the prediction of toxicity for mixtures. Much of the reason may reside in the relative lack of empirical observations and characterizations of chemical interactions. Many QSAR approaches rely on training set approaches to the development of automated programs. Another impediment may be the many examples of the levels, types and biochemical bases for chemical interactions, the intricacies of which would benefit from an automated approach. This area is a useful area for exploration. 

Computational approaches to predicting toxicity are based on expert systems containing empirical or computationally derived rules, typically encoded via stmctural fragments that could potentially cause toxicity. Several databases are available commercially rule-based DEREK and HazardExpert, as well as systems based on automated QSAR—TOPKAT and CASETOX. Computational methods for predicting drug toxicity are reviewed in reference 57. 

TOPKAT (Accelrys Inc.) is a commercially available system and contains a range of cross-validated QSARs, which are multivariate statistical relationships between experimentally derived toxicity data and chemical descriptors that quantify chemical transport properties and biochemical interaction with the target site. It also provides the user with a measure of whether the query molecule fits within the prediction space of the chosen relationship and therefore whether the estimation is rehable. For this comparison exercise, es-... 

---