Immunological variables and tumor types influence one-year survival probability in cancer patients: A comprehensive analysis using logistic regression and decision tree models
Lopez Malizia, A.
Show abstract
The present study aimed to explore immunological variables associated with survival, TP53 gene expression, and primary diagnosis in patients with cancer. Based on these variables, logistic regression and decision tree models (lightGBM) were used to model the probability of one-year survival of patients following their initial diagnosis. Logistic regression revealed the significance of primary diagnosis categories such as Malignant Melanoma, Ovarian Cancer, and Glioblastoma as predictor variables. For the classification model, in addition to these tumor types, variables related to the immune system were also found to be important, including tumor cell percentage, stromal cell percentage, lymphocytes, and necrotic cells. In addition, unsupervised classification techniques were employed to explore the numerical dataset. For this methodology, the best clustering cohesion was observed with two groups determined using different algorithms. The clusters generated by k-means and DBSCAN exhibited differences in the proportion of infiltrating lymphocytes, neutrophils, and monocytes.
Matching journals
The top 5 journals account for 50% of the predicted probability mass.