Abstract: Systems and methods are provided for identifying a set of highly-correlated genes for use in classifying both a first type of biological sample and a second type of biological sample as to a phenotypic characterization, where the first type of biological sample and the second type of biological sample are a sample type selected from the group consisting of a fresh biological sample, a frozen biological sample, a biological sample that has been preserved with a non-crosslinking preservative, and a fixed biological sample that has been fixed with a crosslinking agent, and where the first type of biological sample and the second type of biological sample are sample types that are not the same sample type. The invention also relates to computer systems and methods for use in training a classifier using the highly-correlated genes, and using the trained classifier to classify biological samples that are of the same sample type as the first type of biological sample or the second type of biological sample.