Fig. 1: Illustration of the research framework for data-driven virtual patient generation and in silico clinical trials using the QCIC model.

a Short-term efficacy indices for solid tumors derived from clinical trial reports. b Tumor-immune interaction mechanisms across multiple compartments. The TDLN compartment focuses on antigen presentation and cell differentiation, the PB compartment facilitates immune cell transport between tissues, the TME represents the complex interplay between tumor and immune cells, and the BT compartment serves as the primary site for immune cell production. c The PK/PD model. d Mathematical description of the seven main steps of the QCIC model. Step 1: Tumor cell antigens released via mechanisms including normal apoptosis (\({\kappa }_{1}\overrightarrow{D}\)), immune attack (\({\kappa }_{2}\overrightarrow{I}\)), and chemotherapy-induced killing (\({\kappa }_{3}\overrightarrow{C}\)). Step 2: Tumor antigen presentation is facilitated by activated dendritic cells. Step 3: T-cell priming and activation, the differentiation and proliferation are described by the Michaelis-Menten and Hill functions, respectively. Steps 4 and 5: Immune cell migration and chemotaxis. Step 6: Tumor cell heterogeneity that is represented by drug-sensitive (X₁), drug-resistant (X₂), and drug-pressure (X₃) tumor cell types. Step 7: Tumor-immune interactions that involve immune cell-mediated targeting and tumor-induced immunosuppression via the PD-1/PD-L1 pathway. e Generate virtual patients displaying clinical disease progression by adjusting tumor heterogeneity parameters. f Predicting inter-individual treatment differences in advanced mCRC patients through in silico clinical trials. g Prediction of survival rates for different mCRC patients based on QCIC model-derived predictive biomarkers. For detailed descriptions of mechanisms, symbol meanings, and dynamic equations, refer to Methods 4.2, Supplementary Text 1, and Supplementary Text 2.