This a prospective real-world navigation study using tumor DNA sequencing technology to sequence genes of previously treated and refractory gastrointestinal tumors, which are generally considered to be highly heterogeneous and complex, to screen potential molecular targeted drugs for individualized treatment. This study may provide feasibility and response information, which will be the basis for designing better randomized trials, which may change the pattern of cancer treatment. If the hypothesis is finally proved, it will help doctors and molecular biologists to choose the best drug (or combination of drugs) based on the individual oncogenomics of each patient.

This is a prospective, open label, real-world study to evaluate the feasibility of matched molecular targeted therapy in patients with refractory gastrointestinal cancer based on tumor molecular characteristics as standard treatment failure. This is a non-randomized trial designed to test molecular matching strategies based on patient genome information. The molecular tumor board (MTB) will recommend treatment, but the treatment decision is up to the attending physician. FoundationOne CDx was used for tissue genomic analysis. If possible, PD-L1 immunohistochemistry (IHC), tumor mutation load (TMB), and microsatellite instability (MSI) status will also be evaluated. Based on this information, MTB, composed of multidisciplinary experts, will focus on the selection of customized, best matched drugs or combinations to target most of the genomic changes in each patient, also taking into account potential drug toxicity. The final treatment will be based on the choice of oncologists, who will formulate treatment plan by combining MTB discussion, patient preference, attention to complications, consideration of drug toxicity, insurance coverage of off label drugs and availability of clinical trials of research drugs, thus reflecting the actual clinical practice nowadays in China. The acquisition of drugs follows the actual situation in the real world and is not within the scope of this study design.

The principles of MTB treatment recommendations can be referred to the NCI-MATCH trial:

Grade 1: FDA / NMPA approval; clear evidence and mechanism; concomitant diagnostic relationship between drug and target.

Grade 2: The drug reaches the clinical end point (objective response rate, PFS or OS); there is evidence of target inhibition; there is strong evidence to predict relationship between the target and the drug.

Grade 3: The drug has evidence of clinical activity and target inhibition; there is some evidence to predict the relationship between target and drug.

Grade 4: Pre-clinical evidence of anti-tumor activity and evidence of target inhibition; there is hypothesis to predict the relationship between target and drug.

Patients who meet the inclusion criteria and sign informed consent will be analyzed for tumor mutation by FoundationOne CDx detection. Biopsy specimens of metastatic foci can be selected for gene detection. If there is no targeted therapy among patients, the archived diagnostic tumor samples can also be used for detection. If the patient has received targeted therapy, the investigators will try to obtain biopsy samples after treatment. If the sample cannot be tested or the test results are not satisfactory, additional slices need to be sent to the laboratory. If the repeated test fails, repeat biopsy if the clinical conditions permit, or patients should be included in the unmatched treatment group.

Determining the interventability of gene mutations depends on genomic changes, allele frequency, and diagnosis. This study will combine the recommendations of MTB and gene testing reports to determine the interventability of mutated genes.

Interventional mutation: cancers have FDA or NMPA approved therapies, or there are matching drugs in clinical trials.

Non-interventional mutation: there is no FDA or NMPA approved treatment for cancers, and no matching drug is available in clinical trials.

MTB should at least be composed of attending doctors, molecular pathology experts and bioinformatics experts. The research center is responsible for the establishment and holding of MTB. As management of patients enrolled in this study is essential, the research committee, including but not limited to the main investigator of the study and the treated medical oncologist, can provide advice as needed between MTB meetings to avoid treatment delays. By analyzing the genomic variation of patients, independent MTB recommends molecular targeted therapy that can inhibit the mutation directly or through related pathways or be included in relevant clinical trials. The independent research committee will use the "N-to-1" model of precision medicine to propose a matching therapy consisting of at least one drug that targets molecular changes. The investigators expect 50-70% of patients to carry intervening variants, and that proportion is rising rapidly as learning more about these pathways and drugs. The investigators expect that about 40% of patients with intervening variants will be able to receive molecular targeted "matching" therapy.

Inclusion Criteria:

• Histologically confirmed recurrent or metastatic malignant tumors of digestive tract, including but not limited to:

  1. Biliary tract cancer (including gallbladder cancer and cholangiocarcinoma)
  2. Gastric cancer
  3. Esophageal squamous cell carcinoma
  4. Colorectal cancer
  5. Gastrointestinal stromal tumor
  6. Pancreatic cancer
  7. Primary unknown metastatic carcinoma of digestive system

• failure of conventional treatment;

• have at least one measurable lesion according to RESIST1.1;

• the target lesion is not suitable for local treatment;

• the expected survival time was more than 3 months;

• age ≥ 18 years old;

• the main organs function well;

• be able to swallow and retain oral medication if necessary;

• patients must have enough tissue samples for gene mutation detection;

• informed consent signed.

Exclusion Criteria:

• main lesions were suitable for local treatment;
• serious or uncontrolled medical diseases that researchers consider to be confusing in the treatment response analysis (i.e. uncontrolled diabetes, chronic kidney disease, chronic lung disease or uncontrolled active infection, mental illness / social status that limits compliance with research requirements);
• pregnant or lactating patients or any fertile patients taking no appropriate pregnancy prevention.