BackgroundStage II colorectal cancer(CRC) patients after surgery alone have a five-year survival rate of ∼60-80%; the incremental benefit of adjuvant chemotherapy is <5%. Predicting risk of recurrence and selecting effective personalized adjuvant drugs for stage II CRC using formalin-fixed, paraffin-embedded(FFPE) samples is a major challenge.Methods1319 stage II CRC patients who enrolled in 2011-2019 at Sun Yat-sen University Cancer Center were screened. RNAseq data of FFPE tumor samples of 222 stage II MSS CRC patients(n.recurrence=47, n.norecurrence=175, median follow-up=41 months) were used to develop a method TFunctionalProg for dissecting heterogeneous subgroups of recurrence, predicting risk of recurrence and proposing adjuvant drugs.ResultsTFunctionalProg showed significant predictive values in 222 stage II MSS CRCs. The TFunctionalProg low-risk group had significantly better RFS (validation set (HR=4.78, p-value=1e-4, low-risk group three-year RFS=92.6%, high-risk group three-year RFS=59.7%). TFunctionalProg dissected two subgroups of transition states of stage II MSS CRCs at a high risk of recurrence; each state displays distinct levels of hybrid epithelial-mesenchymal traits, cytotoxic cell inhibition and FOLFOX resistance. Based on mechanisms in two subgroups, TFunctionalProg proposed personalized rational adjuvant drug combinations of immunotherapy, chemotherapy and repurposed CNS drugs. The complementary utility of TFunctionalProg and ctDNA-based prognostic biomarkers were presented.ConclusionTFunctionalProg was validated using FFPE samples to predict the risk of recurrence and propose rational adjuvant drug combinations for stage II CRCs.