NCI Division of Cancer Biology Workshop
September 20-21, 2018
National Institutes of Health, Bethesda, MD
The previous five years have seen a paradigm-shifting ascendance of artificial neural network-based methods in computer vision, imaging, and natural language processing – collectively and popularly called “Deep Learning.” A key methodology within the broader fields of artificial intelligence and machine learning, deep learning approaches have been rapidly adopted for use in health records mining and in cancer imaging, however their use by the broader cancer research community has been less marked. This can be attributed, at least in part, to a limitation frequently referred to as the “black box,” where deep learning models can make correct predictions but without direct association to underlying mechanisms, hindering biological interpretation. This workshop brought together deep learning researchers, cancer systems biologists, and computational biologists to discuss challenges and opportunities to developing interpretable deep learning methods that can be applied to cancer biology investigations in a manner enabling biological interpretation and knowledge generation.