In this week's Nature, an international group of scientists describes a machine learning-based method that uses DNA methylation signatures to classify central nervous system tumors. The classifier was developed using data from roughly 2,800 cancer patients and is able to identify 91 tumor types. When tested on more than 1,000 CNS tumors previously examined using standard methods, it identified misdiagnoses in 12 percent of the cases. To broaden access to their method, the study's authors have created a freely available online classifier tool. They state that they expect the use DNA methylation signatures with other classification approaches will help improve diagnostic accuracy in all cancers. GenomeWeb has more on this study, here.
And in Nature Genetics, an international research team publishes a genome-wide association study identifying a number of new loci associated with stroke. The study, which involved over 521,000 individuals from a range of ancestral backgrounds, uncovered 22 novel loci and demonstrated shared genetic variation with multiple related vascular traits such as blood pressure and venous thromboembolism. The authors note that the discovery of 11 new susceptibility loci indicates mechanisms not previously implicated in stroke pathophysiology, and they provide a framework for prioritization of stroke risk variants and genes for further functional and experimental follow-up. GenomeWeb also covers this study, here.