AI-Based Tool Measures Cancer Aggressiveness

Tumors, which are a leading cause of death, are becoming more complex, making it harder for researchers to predict their aggressiveness and resistance to treatment. The ability to assess how aggressive a tumor is and predict its behavior, including its likelihood to recur, is essential for improving treatment strategies. Now, a new artificial intelligence (AI) tool aims to address this challenge by predicting the aggressiveness of tumors through the identification of specific proteins. This tool, called PROTsi, measures a "stemness index," where a higher score indicates greater aggressiveness and resistance to drugs.

The machine learning model, developed by researchers from the University of São Paulo (São Paulo, Brazi), generates the stemness index based on protein expression data. To develop PROTsi, the researchers relied on datasets from the Clinical Proteomic Tumor Analysis Consortium (CPTAC), which includes data from 11 types of cancer, such as breast, ovarian, lung, kidney, and pancreatic cancers. The tool was designed to predict the aggressiveness of tumors by analyzing proteins linked to tumor progression and their similarity to pluripotent stem cells. As the disease progresses, the stemness index provides insight into how tumor cells are less differentiated, making them more likely to self-renew and become resistant to treatment.

PROTsi was validated across multiple data sets and shown to be effective in distinguishing between different levels of tumor aggressiveness. The research, published in Cell Genomics, confirmed that PROTsi could predict tumor aggressiveness and identify potential targets for new therapies. The tool demonstrated a better performance in differentiating higher-grade tumors, particularly in uterine, pancreatic, and pediatric brain cancers. While the tool works better for some cancers than others, it has the potential to be applied across a wide range of tumors. The researchers plan to continue refining the tool and testing additional computational models to improve predictions and clinical applications.

“We sought to build a model that can be applied to any cancer, but we found that it works better for some than for others. We’re making a data source available for future work,” said Professor Tathiane Malta from the University of São Paulo.

Related Links:
University of São Paulo