Artificial intelligence–based 5‐year survival prediction and prognosis of DNp73 expression in rectal cancer patients

Abstract

Predictive and prognostic markers in oncology are important areas of clinical cancer research and play essential roles for optimal decision making on therapy and assessing treatment efficacy for precision medicine in cancer patients. While DNp73 has been thought as a biomarker in rectal cancer, its clinical significance is still controversial. Exploration on the role of DNp73 protein expression in rectal cancer is a challenging task because of the limitation of human-based pathology analysis of complex information in pathological images. In this paper, we investigated the power of DNp73 expression identified by immunohistochemistry as predictive and prognostic markers in a cohort of 143 rectal cancer patients from the Swedish rectal cancer trial of preoperative radiotherapy using state-of-the-art artificial intelligence (AI) for machine learning and classification. Average validation results show very high accuracy rates (greater than or equal to to 93%) for the 5-year prediction and prognosis of the rectal cancer patients either with or without preoperative radiotherapy. In comparison with the pathology-based analysis, where DNp73 did not provide any survival information (p > 0.05), not only the findings of the AI-based study are clinically significant, but also pave a new direction for the rapid exploration of biomarkers in oncology.