Machine learning facilitates the application of mass spectrometry-based metabolomics to clinical analysis: A review of early diagnosis of high mortality rate cancers.

Early cancer detection is critical to control disease progression. Research of high mortality rate cancers is thereby essential although cancer diagnosis is challenging. Mass spectrometry (MS)-based metabolomics is a great technique to support. Despite MS-based metabolomics being a high-throughput technique, it can be inefficient due to the exponentially increased data volume and complexity of MS data. To demonstrate the real power of the MS-based platform, machine learning (ML) can provide insights from the sea of data. As an emerging technology, it is worth reviewing the research work on how ML coupled with MS-based metabolomics in the past decade. An overview of clinical metabolomics studies is presented with respect to cancer type. The primary purpose of this review is to present what data strategy has been applied previously and discuss the challenges and potential solutions when MS-based metabolomics and ML are combined for early cancer diagnosis. • MS-based metabolomics is a high-throughput technique to discover diagnostic biomarkers for early cancer discovery. • Machine learning facilitates clinical metabolomics due to its efficiency in analyzing a high volume of data. • Machine learning is a promising tool to detect cancer, classify cancer subtypes, and predict the risk of cancer occurrence. • Robust machine learning models for early cancer diagnosis boost the data network effect in MS-based clinical metabolomics. [ABSTRACT FROM AUTHOR]