LAUSR

Background: Human cancers are heterogenous by their cell composition and origination site. Cancer metastasis generates the conundrum of the unknown origin of migrated tumor cells. Tracing tissue of origin and tumor type in primary and metastasized cancer is vital for clinical significance. Researchers have demonstrated the high performance of DNA methylation-based machine learning models tracing the tissue origin of tumor cells. However, previous models were devised based on tissue site instead of tumor type, generating potential problems of indistinguishable tumor subtypes from the same site, e.g., esophageal squamous cell carcinoma versus esophageal adenocarcinoma. Methods: we employed a novel tumor-type-specific hierarchical model using DNA methylation data to develop a multilayer perceptron model (MLP) to trace tumor tissue of origin and subtype. DNA methylation data on 7735 tumor samples from 27 cancer types were downloaded from GEO and TCGA to develop tumor-type-specific libraries. The discovery data set was split into 80% training and 20 % testing. The tumor classifier hierarchy was established with two layers for 27 cancer types. In each layer of the hierarchy, the MLP model was trained using the selected cancer-type discriminatory CpGs. HiTAIC was established based on the four hierarchical MLP models. To validate HiTAIC, we applied the model to the external validation data sets computing stratified and overall precision, recall, and f1-score to evaluate the performance. Next, we applied the model to the application data sets and used stratified and overall precision, recall, and f1-score to evaluate model performance in metastasized cancers. Results: We observed high accuracy of HiTAIC's performance in primary cancer tracing with 99% accuracy and 99% weighted average F1-score in the testing dataset. We observed 93% accuracy and 93% weighted average F1-score for external validation across 25 cancer types. In metastasized cancer, HiTAIC demonstrated 96% accuracy and 98% weighted average F1-score across five cancer types with six different metastatic locations. HiTAIC traces tumor tissue of origin and cancer subtype with high accuracy in primary and metastasized cancer. Conclusion: We developed HiTAIC, a DNA methylation-based classifier, to trace tissue of origin and tumor type in primary and metastasized tumors. The model's capability to trace the tumor origin and subtype with high resolution and accuracy promises potential clinical use in identifying cancer of unknown origin. HiTAIC can be easily deployed in a web-based application, transforming computational sophistication into a user-friendly public tool. Citation Format: Ze Zhang, Brock Christensen, Lucas Salas. HiTAIC: Hierarchical tumor artificial intelligence classifier traces tissue of origin and tumor type in primary and metastasized tumors using DNA methylation data. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 6551.