To further explore the potential of the S-score system to identify genes related to different clinical parameters, breast cancer patients from the TCGA cohort were divided according to two hormonal subtypes: ER+PR+ and ER-PR-. Data from patients in each subtype were then used to calculate the S-scores for all human been previously identified as been involved in the development of the respective tumor types. The S-score also allows for a direct comparison between samples classified differently according to a biological and/or clinical parameter. To illustrate this application, the samples in the TCGA high-grade serous ovarian cancer data were divided into quartiles according to overall survival. We then calculated the Sscore for all human genes using the samples belonging to the first and last quartile of the survival distribution. A comparison of S-scores calculated from the two groups allowed us to identify putative oncogenes and putative tumor suppressor genes associated with either the shortest or the longest survival. Several of the genes identified are known markers for survival. For example, CDC42 inhibition has been associated with longer survival in mice with prostate cancer xenografts. Another example is CANX whose down-regulation has been associated with longer survival in GBM patients. Furthermore, genetic variants of RGS12 have been associated with survival in late-stage non-small cell lung cancer. Another interesting gene is TJP2 whose over-expression has been associated with long-term survival in GBM, in agreement with the pa ern shown in Figure 3. Among the genes identified by this Atractylenolide-III scoring system to be associated with survival, the most interesting are those with opposite classifications in the shortest or the longest survival quartiles. We found that glucoronidase B had a positive score for the shortest survival group and a negative score for the longest survival group. Glucuronidases are known for being involved in the spreading of tumor cells from the primary site and GUSB has been recently included in a signature for predicting lymph node metastasis in cervical cancer. The S-score method confirms the idea that GUSB has an oncogenic function in the more aggressive tumors. However, its negative S-score in the less aggressive tumors indicates that the loss of GUSB might also drive ovarian cancer development with the resulting tumors being less aggressive. An interesting finding in our analysis is the association of RAD23B and XPC, both with negative S-scores, with short-term survival. Proteins encoded by these genes form a complex involved in DNA-damaged repair. A number of other genes with opposite S-scores in the shortest and the longest survival groups are presented in Figure 3. These genes may represent potential prognostic biomarkers as well targets for the development of new therapies. To further explore the potential of the S-score system to identify genes related to different clinical 
parameters, breast cancer patients from the TCGA cohort were divided according to two hormonal subtypes: ER+PR+ and ER-PR-. Data from patients in each subtype were then used to calculate the S-scores for all human genes. While the oncogenes in the two Kaempferide subtypes are basically the same, a much larger discordance is observed for tumor suppressor genes. T