Using machine learning explainer SHAP to improve genomic data heatmap visualisation on a dashboard

Finding insights in large and complex high-dimensional numerical genomic data with many numerical values of the gene expression or protein marker is challenging. Machine learning (ML) explainer SHapley Additive exPlanations (SHAP) is an explainable artificial intelligence (XAI) approach which provides insights into how each feature influences the model's predictions and enhances transparency and trust in ML models. Visualisations are typically used to support XAI, but there is limited research that utilises XAI to enhance visualisation outcomes. This paper presents a novel method that utilises an interactive visualisation dashboard, incorporating explainer SHAP, to select more relevant genomic features and enhance the visual analytics dashboard outcomes. This is crucial for genomic data analytics due to the high number of numerical gene features, which is not effective in analysing and visualising with the limited space and human cognitive load. We demonstrate the effectiveness of our new method on two case studies with Rhabdomyosarcoma (RMS) datasets and Acute Lymphoblastic Leukaemia (ALL) with over 200 selected gene features. The ML explainer SHAP method ranks and finds essential features from a high-dimensional numerical genomic dataset, and the ranked results are applied to the heatmap on the interactive visualisation dashboard to reduce cognitive clutter.