Functional genomics is a typical scientific discovery domain characterized by a very large number of attributes (genes) relative to the number of examples (observations). The danger of data overfitting is crucial in such domains. To avoid this pitfall and achieve predictor robustness, state-of-art approaches construct complex classifiers that combine relatively weak contributions of up to thousands of genes (attributes) to classify a disease. The complexity of such classifiers limits their transparency and consequently the biological insight they can provide. The goal of this study is to apply to this domain the methodology of constructing simple yet robust logic-based classifiers amenable to direct expert interpretation. The approach is based on the subgroup discovery rule learning methodology, enhanced by methods of restricting the hypothesis search space by exploiting the relevancy of features that enter the rule construction process as well as their combinations that form the rules. A multi-class functional genomics problem of classifying fourteen cancer types based on more than 16000 gene expression values is used to illustrate the methodology. Some of the discovered rules allow for novel biological interpretations.