Improving cutaneous melanoma prognosis through adjusting PSMB9 expression: A diagnostic accuracy study based on the construction and optimization of an artificial neural network model.

This study uses the modified artificial neural networks (ANN) model to predict the impact of PSMB9 indicator changes on the prognosis of patients in the death group. Building and training a binary classification ANN based on survival and death groups of patients with cutaneous melanoma (CM), the 10 therapeutic decision biomarkers as input data which were selected by previously our study. Using the death group as the prediction dataset, the expression level of PSMB9 is modified to observe how many samples in the prediction dataset are classified into the survival group or the death group by the ANN model. Four hundred sixty-seven CM patients with different prognoses were included in the TCGA-SKCM dataset, with 10 decision-making treatment biomarkers selected as input data. A binary classification ANN model was built based on the prognosis of CM patients, achieving 100% accuracy. After adjusting the expression of PSMB9 twice, based on the original values, the model predicted that 5 to 35 patients in the prediction dataset would be classified as the survival group. Based on the cutoff value increased to the maximum, the model predicted that 21 to 187 patients in the prediction dataset would be classified as the survival group. Statistical analysis of PSMB9 expression at various stages indicated that its expression decreases with the progression of the T stage, with no differences in the M- and N-stages. Immune infiltration analysis suggested that PSMB9 is also involved in regulating multiple immune cell abundances. The expression of PSMB9 at various stages and the screening of related immune cells were explored. The developed ANN model can predict changes in prognosis based on changes in some indicators in the medical field.