Research Paper Volume 16, Issue 16 pp 11939—11954

Figure 5. Validation of the AML subtype training model. (A) The GSE71014 cohort from the GEO database was downloaded as the internal test group, and the following final model coefficients were constructed: risk score = [value of SOCS1×(0.70)] + [value of HSPA1B×(0.40)] + [value of PLXNB1×(-0.55)] + [value of IL2RA×(0.93)] + [value of INSR×(-1.47)] + [value of UNC93B1×(0.80)]. (B–D) The training cohort, internal test group, and external test cohort were divided into high-risk and low-risk groups. (E, F) The risk scores and survival times of the different patients in the TCGA-LAML cohort and GSE71014 cohort. (G, H) The prognosis of the high-load patients was better than that of the low-load patients due to the presence of more mutation sites. (I) We constructed a column chart to better guide the identification of high-risk patients in the clinic.