Genetic predisposition and misdiagnosis of cancer in All of Us participants

SUMMARY/ABSTRACT The National Cancer Institute estimates ~40% of persons in the United States will be diagnosed with cancer at some point in their lives (https://www.cancer.gov). While 5-year survival rates continue to increase due to improvements in clinical care, the Centers for Disease Control and Prevention reports cancer is still the second

leading cause of death in the United States (www.cdc.gov). Innovative measures and larger datasets are required to continue these improving trends in clinical care(Rahib et al. 2021). The last 15-years of cancer research has benefited tremendously from the advent of next-generation sequence technologies. Ever present

in this genomics revolution is The Cancer Genome Atlas (TCGA). For over a decade, TCGA led the way to molecularly characterize over 10,000 tumors from 33 different cancer types(Ellrott et al. 2018; Ding et al. 2018; Bailey et al. 2018). From these efforts arose a common theme that all tumors are unique, but many share

prognostic and diagnostic drivers of disease. Among these biomarkers are cancer predisposition or germline mutations contributing to cancer development(K.-L. Huang et al. 2018). Despite this large effort, TCGA is a case set heavily biased toward cancer type selection and post-cancer data collection, thus making it difficult to identify

predictive or preventive disease models. To address this issue, and many others concerning human health, the National Institutes of Health has united to produce the All of Us Research Program(Ramirez et al. 2022). This phenomenal program currently has over 400,000 participants who have agreed to share their electronic health records (EHR) and genetic

information(Doerr et al. 2021). This number is expected to grow to one million by its conclusion. Participant selection is disease agnostic, and recruitment has focused on underrepresented minorities, with almost 50% of participants reporting non-White. Preliminary analysis of the insurance billing codes suggests the All of Us

collection will be a fruitful dataset to study cancer. We found 35% (34,849 of 98,553, version 6 release) have (or had) reported neoplasms. Furthermore, this 35% makes up ~80% of all billing code occurrences shared in the electronic health record, again highlighting the All of Us dataset will be a rewarding environment to study cancer.

Here, we propose two ambitious aims run by two teams of undergraduate students that will achieve our overall objective to characterize and quantify the impact of known predisposition cancer mutations and develop models for cancer misdiagnosis in the All of Us Research Program. Separated by genotype and

environment, these aims seek to i) identify and assess the genetic intersection of cancer predisposition databases with the All of Us genomics cohort and ii) discover computational algorithms and features that can predict cancer misdiagnosis. Collectively, these aims encompass doable tasks for well-trained undergraduates

in bioinformatics. We look forward to advancing the cancer research community beyond tumor-specific phenotypes by exploring the whole individual to find novel links to comorbidities and cancer triggers to help elucidate the missing heritability in cancer.