Patient-derived xenografts (PDXs) have become a prominent model for studying human cancer in vivo. The underlying assumption is that PDXs faithfully represent the genomic features of primary tumors, retaining their molecular characteristics throughout propagation. However, the genomic stability of PDXs during passaging has not yet been evaluated systematically. Here we monitored the dynamics of copy number alterations (CNAs) in 1,110 PDX samples across 24 cancer types. We found that new CNAs accumulated quickly, such that within four passages an average of 12% of the genome was affected by newly acquired CNAs. Selection for preexisting minor clones was a major contributor to these changes, leading to both gains and losses of CNAs. The rate of CNA acquisition in PDX models was correlated with the extent of both aneuploidy and genetic heterogeneity observed in primary tumors of the same tissue. However, the specific CNAs acquired during PDX passaging differed from those acquired during tumor evolution in patients, suggesting that PDX tumors are subjected to distinct selection pressures compared to those that exist in human hosts. Specifically, several recurrent CNAs observed in primary tumors gradually disappeared in PDXs, indicating that events undergoing positive selection in humans can become dispensable during propagation in mice. Finally, we found that the genomic stability of PDX models also affected their responses to chemotherapy and targeted drugs. Our findings thus highlight the need to couple the timing of PDX molecular characterization to that of drug testing experiments. These results suggest that while PDX models are powerful tools, they should be used with caution.peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/167767 doi: bioRxiv preprint first posted online Jul. 24, 2017; Cancer research relies on interrogating model systems that mirror the biology of human tumors. Cell lines cultured from human tumors have been the workhorse of cancer research for many years, but the marked differences between the in vitro cell line environment and the in vivo tumor environment raise concerns that cell lines may not be fully representative of human tumors. Recently, there have been increasing efforts to utilize patient-derived xenografts (PDXs) as models to study drug response [1][2][3][4] . These in vivo models are assumed to capture the cellular and molecular characteristics of human cancer better than simpler cancer model systems such as established cell lines grown in vitro and in vivo as xenografts 1,2 . The value of PDX models for cancer research thus depends on their faithfulness in representing the biological features of primary tumors.There are reasons to suspect PDX models might deviate from human tumor biology, as they must be serially transplanted for multiple generations in the murine microenvironment. Therefore, it is important to assess whether PDXs retain their...