The question is: how reliable are the results of animal testing when used to predict effects on humans. After all, a human is quite different from a mouse or a rat.
High doses over a short period
The animal on which a substance is being tested is fed high doses for a short period, whereas the concentrations in our food would be much lower, allowing us to eat the product for our entire life. Furthermore, the tests only examine a limited number of health effects.
Searching for high-risk components
We already know whether a large number of substances, genes and proteins are harmful or not. Analysing the composition of a product is a fast way of identifying potential high-risk components.
You can do this using analytical chemical or molecular biological tools. When assessing the safety of GMOs, for example, the gene expression patterns are compared with similar, safe, non-GMOs. A statistical analysis then shows whether the GMO in question is similar to a standard variety (and therefore safe) or requires more detailed examination.