BackgroundPatient-derived xenograft (PDX) models have shown a great efficiency in preclinical and translational applications. Gastrointestinal (GI) tumors have a strong heterogeneity, and the engraftment rate of PDX models remarkably vary. However, the clinicopathological and molecular characteristics affecting the engraftment rate still remain elusive.MethodsA total of 312 fresh tumor tissue samples from patients with GI cancer were implanted into immunodeficient mice. The median follow-up time of patients was 37 months. Patients’ characteristics were compared in terms of PDX growth and overall survival. PDX models of 3-6 generations were used for drug evaluation.ResultsIn total, 171 (54.8%, 171/312) PDX models were established, including 85 PDX models of colorectal cancer, 21 PDX models of esophageal cancer, and 65 PDX models of gastric cancer. Other than tumor site, histology, differentiation degree, and serum alpha-fetoprotein (AFP) level, no significant differences were found between transplantation of xenografts and patients’ characteristics. For patients who had undergone neoadjuvant therapy, the incidence of tumor formation was higher in those with progressive disease (PD) or stable disease (SD). In gastric cancer, the results showed a higher transplantation rate in deficient mismatch repair (dMMR) tumors, and Ki-67 could be an important factor affecting the engraftment rate. The gene mutation status of RAS and BRAF, two important molecular markers in colorectal cancer, showed a high degree of consistency between patients’ tumors and PDXs. However, no significant effects of these two mutations on PDX engraftment rate were observed. More importantly, in this study although KRAS mutations were detected in two clinical cases, evident tumor inhibition was still observed after cetuximab treatment in both PDX models and patients.ConclusionA large-scale PDX model including 171 cases was successfully established for GI tumors in our center. The relationship between clinicopathological and molecular features and engraftment rates were clarified. Furthermore, this resource provides us with profound insights into tumor heterogeneity, making these models valuable for PDX-guided treatment decisions, and offering the PDX model as a great tool for personalized treatment and translation research.