Multiscale in silico knowledge-based model predicts tumor growth in advanced EGFR+ NSCLC patients on gefitinib.

e20565 Background: Tumor heterogeneity and treatment resistance remains a continued threat to patients with advanced non-small cell lung cancer (NSCLC), despite the emergence of targeted medicines, e.g. against epidermal growth factor receptor (EGFR) alterations. We developed an in silico EGFR mutated lung adenocarcinoma (EGFR+ LUAD) model to predict the effect of known oncogenic EGFR mutations (common EGFR mutations, EGFR exon 20 insertions). The model provides mechanistic representation of tumor progression, including response to gefitinib and captures tumor heterogeneity, patient age, gender, initial clinical disease stage, and smoking status. Methods: 5-step in silico model development: Model building: biology of EGFR+LUAD was characterized by extracting biological features and their functional relationships from > 300 published papers and translating them into ordinary differential equations (ODEs). Mutational burden, EGFR-downstream-pathways, tumor growth and heterogeneity, gefitinib-PK/PD, treatment-induced resistance and clinical outcome were incorporated into a knowledge-based model containing 27-97 variables, 108-258 parameters and 13-83 ODEs reflecting intra-tumor clonal heterogeneity in a mechanistic manner. Calibration: published spheroid, xenograft and clinical data were used for stepwise calibration to find the correct parameter values. Relevant virtual populations (VPOPs) matching real patients baseline characteristics were generated for model benchmarking and validation. Benchmarking against a published data-based model: coverage of experimental interquartile range (IQR) with simulated IQR (precision) assesses model fit with experimental data, coverage of simulated IQR with experimental IQR (overlap) assesses model fit with experimental variability. Validation: a VPOP with comparable baseline characteristics was simulated and results compared against a published patient dataset that wasn’t used in any step during calibration. Matching of prediction over clinical data was done using both coverage and bootstrapped log-rank metrics. Results: Our model provides comparable outputs to the data-based model without having access to the exact original data (our model: precision of 62%, overlap of 91% VS data-based model: precision of 72% and an overlap of 86%). Besides, as a validation criterion, our model also successfully reproduces the time to progression observed in an independent clinical trial (coverage > 99%, negative log-rank tests > 98%). Conclusions: We simulated tumor growth and treatment response in advanced EGFR+ LUAD patients and successfully validated results both against an existing data-based model and a published clinical data. Our model highlights the potential of in silico modeling to better understand complex diseases progression and support efficient development of innovative therapies.