Discovering Metabolic Subtypes in Melanoma

To examine the similarities between patients' diverse metabolic subtypes and the GSE91061 dataset (melanoma dataset undergoing anti-CTLA-4 and anti-PD-1 therapy), we used a subclass mapping technique. When the p value is decreased, the degree of similarity increases. At the same time, we compared the degree of responsiveness between various subtypes and conventional chemotherapeutic agents (cisplatin, vinorelbine, embellin, and pyrimethamine) using the same methodology.
We employed linear discriminant analysis, also abbreviated as LDA, to create a subtype classification feature index so that we could more accurately measure the immunological features of patients who were represented by a variety of sample cohorts. In the TCGA dataset, we employed the LDA model to compute each patient's subtype feature index, and we examined the feature index of each of the distinct subtypes. Within the TCGA dataset, we assessed the characteristics that were linked to prognosis. Firstly, a z-score was done on each individual feature, and Fisher's LDA optimization standard was utilized to specify that the centroids of each group should be as dispersed as possible. It was discovered that a linear combination A maximized the between-class variance of A in comparison to the variance of the within-class measure. The properties of the model allow for the differentiation between samples of various subtypes analyzed.