Oesophageal adenocarcinoma (EAC) incidence rate is increasing in developed countries. The survival for EAC is low, with less than 22% of patients surviving 5 years, primarily due to late-stage diagnosis and the lack of reliable biomarkers to identify high-risk patients. This underscores the need for effective prognostic tools for early detection. In this study we leverage multi-omics data coupled with machine learning approaches to identify prognostic markers for EAC.
Four data modules encompassing clinical, cancer genomic, germline variants and transcriptomics features were collected from 150 EAC cases from the International Cancer Genome Consortium (ICGC) dataset. Various machine learning algorithms (including logistic regression, random forest, xgboost and ensemble) were applied to classify patient’s prognosis of surviving at least 2 years. We assessed the predictive performance of each machine learning approach using single module and all modules. SHAP (Shapley Additive explanations) was used to identify the features that contributed to the predictive performance of the model.
This study demonstrates the potential of integrating multi-omics data with machine learning to predict outcomes for EAC patients. We will provide a summary of our approach to feature selection, model performance and how this work in identifying cross-modal biomarkers for personalised treatment.