Although breast cancer (BC) has a low mortality rate relative to other cancers, it prominently affects the survival of patients with human epidermal growth factor receptor-2 (HER2 +) BC due to its high recurrence rate. By far, it has been found that autophagy can affect various tumor occurrence and development, as well as patients’ prognosis. HER2 + BC patient samples and autophagy-related genes (ARGs) were acquired from a public database, least absolute shrinkage and selection operator (LASSO) and Cox analyses (including univariate and multivariate analyses) were utilized to construct a 9-ARGs model, which was verified by using HER2 + BC patient samples in The Cancer Genome Atlas (TCGA) dataset. Sample risk score was worked out based on characteristic genes, and prominent differences in overall survival were tracked down between high- and low-risk groups. Predictive ability of the model was validated by drawing receiver operating characteristic (ROC) curves and then calculating the area under the curves (AUC) value. Results showed good accuracy and prediction ability of the model in both validation set and training set. For the purpose of facilitating model application in clinical practice, we constructed a nomogram combing clinical factors and risk scores to evaluate 1-year, 3-year and 5-year survival of HER2 + BC patients. In addition, we assessed the correlation of risk score with tumor mutational burden and tumor immune infiltration. Results exhibited that in a high-risk group, tumor mutation was relatively high, while tumor immune infiltration was relatively poor. Overall, based on ARGs, the prognostic signature in this study can tellingly evaluate prognoses of HER2 + BC patients and provide a reference for clinicians.
