Predictive capacity

Given the absence of correlation between the sensitivity and range descriptors, we also examined whether a two-variable equation would improve on Eq. (2). As shown by Eq. (3), the inclusion of two independent variables in the same equations improved their predictive capacity ... 

Reach target physicochemical profile HT physicochemical profiling and virtual screening Limited relevance of some HT techniques dubious predictive capacity of some statistical models... 

Reach target metabolic behavior, in particular target rate of activation Fast in vitro metabolic assessment and use of expert systems Dubious extrapolation to in vivo situations limited predictive capacity of some expert systems... 

Such essential limitations may markedly decrease the reliability and predictive capacity of quantitative structure-toxicity relationships (STRs) in haloalkenes and all other classes of toxic xenobiotics, but recognition of limitations does not suppress the need for predictive tools. In fact, any approach, empirical or mechanistic, that is able to uncover qualitative STR trends and to assign a priori labels of potential toxicity is certainly welcome. 

The body of samples selected is split into two subsets, namely the calibration set and the validation set. The former is used to construct the calibration model and the latter to assess its predictive capacity. A number of procedures for selecting the samples to be included in each subset have been reported. Most have been applied to situations of uncontrolled variability spanning much wider ranges than those typically encountered in the pharmaceutical field. One especially effective procedure is that involving the selection of as many samples as required to span the desired calibration range and encompassing the whole possible spectral variability (i.e. the contribution of physical properties). The choice can be made based on a plot of PCA scores obtained from all the samples. 

The model can lose its predictive capacity for reasons such as a dramatic change in the nature of the samples. A new calibration model better adjnsted to the new situation must therefore be constructed. In other cases, however, the model can be re-adjnsted (restandardized) in order to ensure a predictive capacity not significantly different from the original one. 

Table 3.6 Comparison of predictive capacities of various equations in estimating standard molar entropy T = 298.15 K, P = 1 bar). Column I = simple summation of standard molar entropies of constituent oxides. Column II = equation 3.86. Column III = equation 3.86 with procedure of Holland (1989). Column IV = equation 3.85. Values are in J/(mole X K). Lower part of table exchange reactions adopted with equation 3.85 (from Helgeson et al., 1978) and Sj finite differences for structural oxides (Holland, 1989). 

ECVAM is the leading international center for alternative test method validation. Hartung et al. (29) summarized the modular steps necessary to accomplish stage 3 (test validation). The seven modular steps are (I) test definition, (2) within-laboratory variability, (3) transferability, (4) between-laboratory variability, (5) predictive capacity, (6) applicability domain, and (7) performance standards (29). Steps 2-4 evaluate the test s reliability steps 5 and 6 evaluate the relevance of the test. Successful completion of all seven steps is necessary to proceed to stage 4 (independent assessment or peer review). This modular approach allows flexibility for the validation process where information on the test method can be gathered either prospectively or retrospectively. The approach is applicable not only to in vitro test methods but also to in silico approaches (e.g., computer-based approaches such as quantitative structure-activity relationships or QSAR) and pattern-based systems (e.g., genomics and proteomics). 

Antman developed a thrombosis in myocardial infarction (TIMI) risk score based on a database of 15,078 patients with STEMI or new onset of complete left bundle branch block (8), The score was validated in the TIMI 9 data set. Ten characteristics of these patients accounted for 97% of the predictive capacity of their multivariate model. These are included in the risk score (Table I). Points were given for difference parameters as listed in Table I. The risk score had a strong association with 30-day mortality. There was a greater >40-fold increase in mortality from TIMI risk score 0 to >8 at 30 days (Table I) (8), The TIMI risk score is easy to apply and can be done at the bedside. 

The CRP levels were determined in these patients. It was higher in patients with plaque rupture (21 patients) compared with those without plaque rupture (24 patients)—3.1 0.5 mg/L versus 1.9 0.4mg/L (P = 0.04). This suggests that an elevated CRP reflects an inflammatory process that can lead to plaque rupture. The CRP seems to have independent predictive capacity for identifying inflammation here. There is also a possibility that CRP may release factors that further weaken the plaque s fibrous cap and allow rupture or erosion (16). 

Appropriate measures of goodness-of-fit, robustness, and high predictive capacity The predictive capacity of the model should be determined by external validation. Further work is recommended to determine what constitutes external validation of (Q)SARs. 

The proposed protocol is a logical extension of the evaluation scheme provided by Plackett and Hewlett (1952) for the problem of predicting effects of complex mixtures of known composition. However, there are as yet only few data available to verify its predictive capacity (Altenburger et al. 2005). Also, ecotoxicity databases usually do not contain the full concentration effect data, which means that reevaluation of existing studies will be laborious. 

The SSD example shows that one model, the SSD, can be considered sufficiently predictive for simple questions (e.g., Is the HC5 protective of community responses ) but of more limited — or even insufficient — predictive capacity for specific assessment questions such as those regarding the effects of exposure on biodiversity. This shows that general conclusions on the validity of a model cannot be drawn. For the extrapolation methods for which validation studies have been done, the examples (and limitations) have been provided in the preceding chapters. 

A method for evaluation of the tiered testing strategy in terms of its predictive capacity and its ability to reduce and refine the use of laboratory animals... 

In this step, one or more independent experts should evaluate the quality of the training set data along with any other available data for the endpoint predicted by the QSAR. This should enable an evaluation to be made of the maximal predictive capacity that could be expected for the QSAR. For QSARs, the inevitable variability in descriptor and response variable data should be taken into consideration when defining criteria for predictive capacity. For example, in the case of a regression-based QSAR, it might be decided that its predictions should fall within a specified prediction interval, and that the R2 value for predictions of independent data should exceed a specified value. Issues relating to the quality of data for use in QSARs are discussed in Cronin and Schultz (2003) and Schultz and Cronin (2003). 

---