Explainable artificial intelligence for multi-omics data.

Abstract

The integration of multiple omics, known as multi-omics, is growing rapidly for developing machine learning (ML) models for biomedical predictions due to the recent advent in next-generation sequencing, clinical investigations, and computing technologies, along with deep learning models for handling high-dimensional features. The multi-omics data allows the models to find complex patterns from the complementary aspects of the data. But, the large number of features in multi-omics data and the black-box nature of ML models induce a lack of interpretability, which can be tackled using eXplainable Artificial Intelligence (XAI) approaches. XAI provides explanations to the model predictions, which impart transparency and a sense of trustworthiness. This chapter discusses the different XAI algorithms and XAI models for multi-omics-based biomedical prediction tasks. In summary, the multi-omics XAI models are crucial for biomedical prediction tasks as multiple omics provide a holistic understanding of the biomedical processes, and XAI imparts interpretability to the ML-based predictions.