The proper orthogonal decomposition (POD) technique (also known as the Karhunen-Loeve transform) has been used as a model reduction tool for many applications in engineering and science. In principle, one begins with an ensemble of data, called snapshots, collected from an experiment or laboratory results. The POD technique is then used to produce a set of basis elements that can span the original snapshot collection using the fewest possible degrees of freedom. It is such capability that allows us to extract the representative characteristics of a cancer from a collection of DNA microarray samples known to be cancerous. The resulting few POD elements can be regarded as dominant cancerous patterns, which can be used to determine whether an arbitrary DNA microarray sample is cancerous. In our study, we consider two types of cancers, liver and bladder. DNA microarray data are downloaded from the Stanford Microarray Database. Our findings indicate that the POD method can successfully detect both cancer types, although our approach can be applied to other types of disease or cancer.