Researchers identify genetic and machine learning tools for cancer risk prediction

Scientists have developed new methods to predict cancer progression and risk using genetic markers and machine learning.

A study published in Science Advances by the Center for Genomic Regulation (CRG) in Barcelona identified a six-gene signature, termed the ‘PHFScore’, to predict survival in patients with acute myeloid leukemia (AML). By analyzing the joint activity of genes including CARMIL1, HOPX, ALDH1A1, CALR, STAR, and PHF19, researchers could categorize patients into distinct survival groups. This scoring system may lead to more affordable and efficient prognostic tests using existing bone marrow samples.

In a separate international study presented at the ESMO 2026 congress, researchers utilized machine learning models to predict the risk of 20 common cancer types. Using data from over 463,000 participants in the UK Biobank followed over 18 years, the models incorporated routine clinical data such as blood tests, medical history, and lifestyle factors. The models, validated against datasets from Scotland, Wales, the United States, and Japan, outperformed traditional statistical methods in identifying individuals at higher risk.