Abstract: Provided herein are methods of predicting cytogenetic abnormalities associated with a cancer in a subject, for example, multiple myeloma. A cytogenetic abnormalities model of a set of reference values obtained from an average of gene expression profile values based on copy number-sensitive genes that correlate to cytogenetic abnormalities associated with the cancer is utilized as a predictive tool. The cytogenetic abnormalities model, as a virtual model (i.e. a “virtual karyotype”), may be tangibly stored with program instructions to implement the model in a computer system. In particular embodiments, the methods and systems provided by the invention operate without FISH (fluorescent in situ hybridization).