Artificial intelligence and deep learning in molecular testing

The sheer amount of information available from increasingly complex data modalities has increased the potential for discovering novel therapeutics and decision aids that can assist with risk stratification. However, these omics and imaging datasets now contain millions of potential variables. As such, this plethora of available information (imaging, omics data) can make it difficult to decide which variables and combinations of such are important for clinical decision making. Machine learning approaches hunt for important patterns in data with many variables and can potentially reduce the burden of sifting through data for key insights. In this book chapter, we introduce machine learning technologies and their potential applications in molecular pathology. This book chapter was written for pathologists with an introductory coding background. For pathologists with no prior coding experience, we have included reference links to introductory programming material. We have also included in the Appendix an in-depth coding tutorial for those seeking a deeper explanation of the implementation of machine learning technologies.