Pan-cancer analysis was used to assess similarities and differences in risk score models between different tumor types. We systematically evaluated TMB, MSI as well as the expression of CD274 among pan-cancer. The risk score was proactively correlated with TMB in BRCA, COAD, LGG, PAAD, STAD and THYM (P < 0.05), while the inverse correlation with TMB in KIRC, KIRP, LAML and UVM (P < 0.05) (Fig.
8A). For MSI, a positive correlation in STAD, DLBC, COAD, HNSC and THCA as well as a negative correlation in CHOL and KIRC, was defined (P < 0.05) (Fig.
8B). Additionally, the risk score was positively relevant to CD274 expression in ACC, COAD, HNSC, LGG, SKCM, THCA and negatively relevant with CD274 content in BRCA, CESC, HNSC, KIRC, LAML, LUSC, OV and PCPG(
P < 0.05) (Fig.
8C). In addition, the mutual relationship between risk score and several immune cell infiltration as well as stemness indices were calculated, respectively (Additional file
6: Fig S6 and (Additional file
7: Fig. S7, (Additional file
8: Fig. S8). Using the immunotherapy cohort of advanced urothelial cancer (IMvigor210 cohort) to evaluate the impact of risk scores on predicting immunotherapy sensitivity. The Log-rank test also showed that GBM patients with a high-risk score were associated with poorer survival conditions (Fig.
8D). Then, integration risk scores were analyzed with immune checkpoint blockade (ICB) treatment studies. The results showed that the proportion of GBM patients of the high-risk score group in the response groups (CR and PR) was notably lower than in the low-risk score group, while the percentage of patients in the no/limited response groups (SD and PD) showed the contrary phenomena, pointing that the risk score could prove the response of GBM patients to ICB therapy. However, in exploring immune phenotypes in high- and low-risk scores, the desert phenomenon was more notable in the low-risk score, while the inflamed was seen more in the high-risk group (Fig.
8E). From combined risk score and tumor neoantigen burden correlation analysis, the GBM patients with low-risk scores together with a high neoantigen burden exhibited the most stretched-out survival time and the GBM patients with high-risk scores in connection with a low neoantigen burden had the worst survival situation (Fig.
8F). The Spearman correlation analysis was shafted to measure the value of the risk score to anticipate drug sensitivity for multiple types of cancer. Finally, 119 drugs for which the risk score and drug sensitivity significantly correlated, were obtained from the GDSC database. Subsequently, we selected the 50 most representative drugs for mapping. The risk score was the most significantly negatively sensitive to 5 drugs, including AZD5991, YK.4.279, Alisertib, Vinblastine and Eg5_9814; and the most significantly positively correlated with sensitivity to 5 drugs, including LJI308, AMG.319, CZC24832, PLX.4720 and PFI3 (Fig.
8G). Among them, the strongest drug sensitivity is LJI308. The study shows that LJI308 is a powerful selective inhibitor of RSK, which can inhibit the growth and proliferation of cancer stem cells [
36]. In addition, the signal path targeted by the selected drugs was discovered. The relationship between drug sensitivity and risk score targeted the cell cycle, mitosis, microtubule, DNA replication and apoptosis regulation signaling were positive. On the contrary, drug sensitivity with negatively related to the risk score targeting the PI3K/mTOR signaling and chromatin histone methylation (Fig.
8H). In summary, the establishment of a risk score will be beneficial in exploring the facility and effective treatment strategies for GBM.