Verily built the DeepMass platform in collaboration with sister company Google and the Max Planck Institute of Biochemistry in Berlin. DeepMass applies deep learning methods to mass spectrometry to detect the presence and levels of proteins associated with various diseases.
While previous mass spectrometry models must be trained on reference material derived from physical experiments, DeepMass’ spectral libraries were built using computation. In the study, this platform proved to be “significantly better” at predicting the presence of those proteins than the previous models, and achieved a level of accuracy close to the theoretical upper limit.
Additionally, when the tool was used to analyze actual clinical data, “we were able to expand the coverage of known biomarkers by more than twofold,” the researchers wrote in a blog post. Of those biomarkers, they added, “The more we know about them and their relationship to specific diseases, the earlier and more precisely we can intervene.”
More articles about AI:
Physician viewpoint: AI will reduce healthcare costs, improve physician wellbeing
NIH’s updated roadmap for AI in radiology focuses on clinical use
Algorithm predicts ICU survival rates more precisely than previous models
