M.S. Candidate: Kamran Karimov
Program: Bioinformatics
Date: 18.01.2024 / 11:00
Place: A-108

Abstract: Cancers of Unknown Primary (CUP) origin are metastases where the primary source of the tumor cannot be detected and only the secondary tumor is evident. This can cause problems in treatment since the tissue of origin defines the base characteristics of the tumor and most therapeutic methods are specific to these characteristics. We built three machine learning models to predict the primary tissue of CUPs based on gene expression profile similarities between the primary tumor and its metastases. The models are trained on 8798 cancer cases across 14 different cancer types obtained from the TCGA program. The specific cancer types are annotated in the data used. During the process, we tried origin prediction based on specific gene types and compared these results with each other and with overall accuracy. The trained model can assist in CUP diagnoses, by further development, using more data.