Back

Multi-Cancer PRS Constellation Model for Cancer Risk Prediction

Moragas, N.; Diez-Villanueva, A.; Moratalla-Navarro, F.; Fernandez-Navarro, P.; Perez-Gomez, B.; Morales Suarez-Varela, M.; Molina-Barcelo, A.; Castano-Vinyals, G.; Rius Sansalvador, B.; Riobo-Mayo, L.; Olmedo-Requena, R.; Jimenez-Moleon, J.-J.; Marcos-Gragera, R.; Guevara, M.; Fernandez-Tardon, G.; Amiano Exezarreta, P.; Huerta, J. M.; Fernandez-Villa, T.; Molina de la Torre, A. J.; Martin-Sanchez, V.; Gomez-Acebo, I.; Dierssen, T.; Alguacil, J.; Guino, E.; Kogevinas, M.; Pollan, M.; Obon-Santacana, M.; Moreno, V.

2024-10-18 genetic and genomic medicine
10.1101/2024.10.17.24315686 medRxiv
Show abstract

Cancer development is influenced by genetic factors and modifiable exposures. GWAS has identified genetic variants and developed of prediction models through Polygenic Risk Scores (PRS), but PRS alone has limitations for estimating cancer risk. This study assesses a novel PRS constellation approach that integrates Polygenic Risk Scores (PRS) from both lifestyle and genetic traits to enhance prediction models for colorectal, breast, and prostate cancers. The approach was developed using the UK Biobank dataset and validated in the independent GenRisk cohort. The model, incorporating sex and age, achieved AUCs of 0.74 for CRC, 0.65 for BC, and 0.75 for PC in the UK Biobank. Including tumor-related PRSs improved PC prediction but had limited impact on CRC and BC. Age and sex inclusion boosted CRC and PC model accuracy. However, GenRisk validation showed reduced AUCs and limited utility of lifestyle PRSs, with CRC and BC models achieving 0.62 and PC 0.56. Integrating lifestyle-related characteristics into PRS does not significantly enhance cancer-specific PRS prediction. However, PRSs for these traits show independent predictive power, highlighting the importance of considering lifestyle in cancer risk and the need for precision medicine to improve early detection.

Matching journals

The top 8 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.