Current orthotopic xenograft models of colorectal cancer (CRC) require survival surgery and do not robustly form tumors in liver, the most common site of metastasis in patients. In the work described in the thesis chapter 2, we used chemokine-targeting to develop cell line and primary patient-derived xenograft models that recapitulate the vast majority of common human somatic CRC mutations as primary gastrointestinal (GI) tumors in mice without requiring surgery. Importantly, we utilize early-stage mouse blastocyst microinjection techniques to extend this approach and model primary human CRCs in immunoproficient mouse hosts. Next, we show that primary GI tumors can inducibly and robustly metastasize to liver. Finally, we demonstrate that human CRC liver metastases in vivo have higher levels of DKK4 and NOTCH signaling and are more chemoresistant than paired sub-cutaneous xenografts. Overall, we anticipate that this experimental system can help improve our mechanistic understanding of human primary CRC progression to liver metastasis and provide a more physiological model than sub-cutaneous xenografts for pre-clinical drug screening. Refined cancer models are urgent to bridge the gap between cell-line or animal based research and clinical research. In thesis chapter 3, we described an organotypic colon cancer model which was generated from human native matrix and have 3 pathophysiologically recapitulated the natural features of progression from APCdependent in situ neoplasia to sub-mucosal invasive adenoma in colorectal cancer (CRC)-associated genetic pathways. To identify invasion-driver genes, we performed a forward genetic screen using Sleeping Beauty (SB) transposon- based mutagenesis in the ex vivo CRC model. This screen identified 39 candidate genes, all of which are listed in TCGA CRC dataset. 17 of them, including TCF7L2, TWIST2, MSH2, DCC and EPHB1,2, most likely drive invasion of CRC through cooperation with mutant APC. Among the remaining genes that have not previously been implicated in CRC, seven out of ten were functionally validated to significantly promote the growth, migration or invasion of colon cells. This piece of work demonstrated the utility of ex vivo humanoriginated models with transposon-based mutagenesis and provided a new system for studying the biology of cancer. 4