Databases comparative QSAR

Comparative QSAR is a field currently under development by several groups. Large databases of known QSAR and 3D QSAR results have been compiled. Such a database can be used for more than simply obtaining literature citations. The analysis of multiple results for the same or similar systems can yield a general understanding of the related chemistry as well as providing a good comparison of techniques. 

In a study by Andersson et al. [30], the possibilities to use quantitative structure-activity relationship (QSAR) models to predict physical chemical and ecotoxico-logical properties of approximately 200 different plastic additives have been assessed. Physical chemical properties were predicted with the U.S. Environmental Protection Agency Estimation Program Interface (EPI) Suite, Version 3.20. Aquatic ecotoxicity data were calculated by QSAR models in the Toxicity Estimation Software Tool (T.E.S.T.), version 3.3, from U.S. Environmental Protection Agency, as described by Rahmberg et al. [31]. To evaluate the applicability of the QSAR-based characterization factors, they were compared to experiment-based characterization factors for the same substances taken from the USEtox organics database [32], This was done for 39 plastic additives for which experiment-based characterization factors were already available. 

Volume 2 of "Exploring QSAR" by Hansch and Leo (1995) contains the most extensive hard-copy database of log Poct values — about 18,000 values for over 16,500 structures. Sangster (1989) has published log P measurements and offers a comparable computer database on diskettes. Several of the computerized calculation programs access a similar log Poct database. If values in other solvent/ water systems are needed, the MASTERFILE database can accompany the CLOGP program on VAX and UNIX systems and contains about 40,000 measured log P values in over 300 solvent systems in addition to about 10,000 measured values of pKa. 

Methods to predict the hydrolysis rates of organic compounds for use in the environmental assessment of pollutants have not advanced significantly since the first edition of the Lyman Handbook (Lyman et al., 1982). Two approaches have been used extensively to obtain estimates of hydrolytic rate constants for use in environmental systems. The first and potentially more precise method is to apply quantitative structure/activity relationships (QSARs). To develop such predictive methods, one needs a set of rate constants for a series of compounds that have systematic variations in structure and a database of molecular descriptors related to the substituents on the reactant molecule. The second and more widely used method is to compare the target compound with an analogous compound or compounds containing similar functional groups and structure, to obtain a less quantitative estimate of the rate constant. 

Potter and Matter [64] compared maximum dissimilarity methods and hierarchical clustering with random methods for designing compound subsets. The compound selection methods were applied to a database of 1283 compounds extracted from the IndexChemicus 1993 database that contain 55 biological activity classes. A second database consisted of 334 compounds from 11 different QSAR target series. They compared the distribution of actives in randomly chosen subsets with the rationally... 

The SYBYL molecular modeling package is known for having the comparative molecular field analysis method for finding three-dimensional QSAR. 2 CJACS hits for CoMFA (Comparative Molecular Field Analysis) are plotted in Figure 20. CoMFA was mentioned in the CJACS database 9% as often as SYBYL, indicating that SYBYL is being used for many functions other than CoMFA. 

A9.6.4.3 When two or more QSARs are applicable or appear to be applicable, it is useful to compare the predictions of these various models in the same way that predicted data should be compared with measured (as discussed above). If there is no discrepancy between these models, the result provides encouragement of the validity of the predictions. Of course, it may also mean that the models were all developed using data on similar compounds and statistical methods. On the other hand, if the predictions are quite different, this result needs to be examined further. There is always the possibility that none of the models used provides a valid prediction. As a first step, the structures and properties of the chemicals used to derive each of the predictive models should be examined to determine if any models are based upon chemicals similar in both of these respects to the one for which a prediction is needed. If one data set contains such an appropriate analogue used to derive the model, the measured value in the database for that compound vs model prediction should be tested. If the results fit well with the overall model, it is likely the most reliable one to use. Likewise, if none of the models contain test data for such an analogue, testing of the chemical in question is recommended. 

Validation is one of the most difficult aspects of environmental QSAR development due to the comparatively small size of the database. Cross-validation has been useful in validating the effectiveness of the model. In this method, one compound is removed from the database, the equation is recalculated, and the toxicity of the omitted compound is estimated. The process is repeated for all compounds in the dataset and the results are tabulated. In this manner, a calculation of the accuracy of prediction of continuous data and the rate of misclassification for categorial data can be compiled. A more useful estimate of the validity of the QSAR model is its ability to predict the toxicity of new compounds. Generally, this is difficult to accomplish in a statistically significant way due to the slow accumulation of new data that meet the criteria used in the modeling process and the associated expense. 

Further additions to the 3D-QSAR arsenal include comparative molecular similarity indices analysis (CoMSIA) [15], 4D-QSAR [16], COMPASS [17], receptor surface models [18], the pseudoreceptor approach [19], ComPharm [20], and comparative molecular surface analysis (CoMSA) [21], 3-D-invariant, alignment-free descriptor systems such as comparative molecular moment analysis (CoMMA) [22], EVA [23], WHIM [24], and ALMOND [25], have also become available. A survey of the 3D-QSAR literature reveals 1154 entries in the Chemical Abstracts Plus database of these, 79% are journal publications, 19% are conference proceedings, and four are patents related to, or using, 3D-QSAR models. As the number of potential targets amenable to drug discovery is increasing exponentially, it is likely that 3D-QSAR models and methodologies will continue to be developed in the next decade. 

---