Bayesian risk prediction model for colorectal cancer mortality through integration of clinicopathologic and genomic data

Melissa Zhao; Mai Chan Lau; Koichiro Haruki; Juha P Väyrynen; Carino Gurjao; Sara A Väyrynen; Andressa Dias Costa; Jennifer Borowsky; Kenji Fujiyoshi; Kota Arima; Tsuyoshi Hamada; Jochen K Lennerz; Charles S Fuchs; Reiko Nishihara; Andrew T Chan; Kimmie Ng; Xuehong Zhang; Jeffrey A Meyerhardt; Mingyang Song; Molin Wang; Marios Giannakis; Jonathan A Nowak; Kun-Hsing Yu; Tomotaka Ugai; Shuji Ogino

doi:10.1038/s41698-023-00406-8

Bayesian risk prediction model for colorectal cancer mortality through integration of clinicopathologic and genomic data

NPJ Precis Oncol. 2023 Jun 10;7(1):57. doi: 10.1038/s41698-023-00406-8.

Authors

Melissa Zhao¹, Mai Chan Lau², Koichiro Haruki², Juha P Väyrynen^{2

3

4}, Carino Gurjao^{2

5}, Sara A Väyrynen^{2

3}, Andressa Dias Costa³, Jennifer Borowsky^{2

6}, Kenji Fujiyoshi², Kota Arima², Tsuyoshi Hamada², Jochen K Lennerz⁶, Charles S Fuchs⁷, Reiko Nishihara^{2

8

9}, Andrew T Chan^{10

11

12

13}, Kimmie Ng³, Xuehong Zhang¹², Jeffrey A Meyerhardt³, Mingyang Song^{9

10

11}, Molin Wang^{8

14}, Marios Giannakis^{3

5

15}, Jonathan A Nowak^#², Kun-Hsing Yu^#¹⁶, Tomotaka Ugai^#², Shuji Ogino^#^{17

18

19

20}

Affiliations

¹ Program in MPE Molecular Pathological Epidemiology, Department of Pathology, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA. mzhao11@bwh.harvard.edu.
² Program in MPE Molecular Pathological Epidemiology, Department of Pathology, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA.
³ Department of Medical Oncology, Dana-Farber Cancer Institute and Harvard Medical School, Boston, MA, USA.
⁴ Cancer and Translational Medicine Research Unit, Medical Research Center Oulu, Oulu University Hospital, and University of Oulu, Oulu, Finland.
⁵ Broad Institute of MIT and Harvard, Cambridge, MA, USA.
⁶ Department of Pathology, Center for Integrated Diagnostics, Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA.
⁷ Genentech/Roche, South San Francisco, CA, USA.
⁸ Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, MA, USA.
⁹ Department of Nutrition, Harvard T.H. Chan School of Public Health, Boston, MA, USA.
¹⁰ Clinical and Translational Epidemiology Unit, Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA.
¹¹ Division of Gastroenterology, Massachusetts General Hospital, Boston, MA, USA.
¹² Channing Division of Network Medicine, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA.
¹³ Department of Immunology and Infectious Diseases, Harvard T.H. Chan School of Public Health, Boston, MA, USA.
¹⁴ Department of Biostatistics, Harvard T.H. Chan School of Public Health, Boston, MA, USA.
¹⁵ Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA.
¹⁶ Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA.
¹⁷ Program in MPE Molecular Pathological Epidemiology, Department of Pathology, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA. sogino@bwh.harvard.edu.
¹⁸ Broad Institute of MIT and Harvard, Cambridge, MA, USA. sogino@bwh.harvard.edu.
¹⁹ Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, MA, USA. sogino@bwh.harvard.edu.
²⁰ Cancer Immunology and Cancer Epidemiology Programs, Dana-Farber Harvard Cancer Center, Boston, MA, USA. sogino@bwh.harvard.edu.

^# Contributed equally.

Abstract

Routine tumor-node-metastasis (TNM) staging of colorectal cancer is imperfect in predicting survival due to tumor pathobiological heterogeneity and imprecise assessment of tumor spread. We leveraged Bayesian additive regression trees (BART), a statistical learning technique, to comprehensively analyze patient-specific tumor characteristics for the improvement of prognostic prediction. Of 75 clinicopathologic, immune, microbial, and genomic variables in 815 stage II-III patients within two U.S.-wide prospective cohort studies, the BART risk model identified seven stable survival predictors. Risk stratifications (low risk, intermediate risk, and high risk) based on model-predicted survival were statistically significant (hazard ratios 0.19-0.45, vs. higher risk; P < 0.0001) and could be externally validated using The Cancer Genome Atlas (TCGA) data (P = 0.0004). BART demonstrated model flexibility, interpretability, and comparable or superior performance to other machine-learning models. Integrated bioinformatic analyses using BART with tumor-specific factors can robustly stratify colorectal cancer patients into prognostic groups and be readily applied to clinical oncology practice.

Abstract

Grants and funding