Background: Inevitable progression on 3^rd-generation EGFR-tyrosine kinase inhibitor (TKI) osimertinib of EGFR-mutated lung cancer patients represents a great challenge in clinic. Previous studies have revealed that one-third of the resistant mechanisms are due to acquired EGFR secondary mutations, mainly on C797, L718 and L792 residues. Our study aims to gain insights into novel mechanisms of acquired resistance to osimertinib. Methods: We performed genomic profiling on a total of 1,058 EGFR-mutated lung cancer patients with progressed disease on osimertinib, and a cohort of 1,803 patients who received only 1^st-generation EGFR TKIs upon progression. Recurrent EGFR mutations with a significant enrichment in the osimertinib group were identified. We further established and applied molecular dynamic simulation-based computational model of the mutant EGFR protein to predict its sensitivity to osimertinib. Results: As expected, compared with 1^st-TKIs alone group, EGFR mutations, including C797S/G (22.1% vs. 0.5%), L718Q/V (6.2% vs. 0.3%), L792F/H (4.4% vs. 0.3%), were significantly more enriched in the osimertinib cohort. Our computational model has also successfully predicted their sensitivities to osimertinib: WT (-35.19 kcal/mol) > L792F (-34.10 kcal/mol) > L718Q (-30.33kcal/mol) > C797S (-28.02 kcal/mol), which are consistent with our previous in vitro validations. Importantly, a total of 14 low-frequency EGFR mutations were exclusively observed in the osimertinib group, seven of which, including EGFR G796S(n = 6), V802F(n = 3), T725M(n = 2), Q791L/H(n = 2), P794S/R(n = 2), were predicted to dramatically reduce the binding affinity of osimertinib to EGFR. Of note, analysis of the pretreatment samples of two patients supported that EGFR V802F and Q791L/H were acquired during osimertinib treatment. Interestingly, EGFR G796S was predicted to be sensitive to gefitinib, suggesting the possibility of administration of gefitinib in patients with acquired EGFR G796S to first-line osimertinib treatment. Further in vitro functional validations are currently ongoing. Conclusions: Our study represents the largest EGFR-mutated lung cancer cohort so far to investigate osimertinib resistance in a real-world setting, and has uncovered a list of recurrent low-frequency EGFR mutations that may confer acquired resistance to osimertinib. Our in silico structural model was proved to be powerful and robust in the prediction of osimertinib sensitivity of EGFR mutants.