A new artificial intelligence system can top human experts in breast cancer prediction, according to a report in Nature this week. The deep learning AI system works by screening mammograms and was assessed using two datasets: one from the UK consisting of 25,856 mammograms and one from the US consisting of 3,097 mammograms. The Google Health-led team that developed the system demonstrates an absolute reduction of 5.7 percent and 1.2 percent (US and UK) in false positives and 9.4 percent and 2.7 percent in false negatives. The system also outperformed six radiologists in an independent follow-up study, and was found to significantly reduce the workload of the second reader in the mammogram double-reading process used in the UK.
A universal sample multiplexing method for single-cell RNA sequencing was described recently in Nature Biotechnology. The approach involves chemically labeling fixed cells by attaching DNA oligonucleotide tags to cellular proteins. In a 96-plex perturbation experiment, the technique enabled the identification of changes in cell population structure and transcriptional states that cannot be discerned from bulk measurements, establishing it as "an efficient method for surveying cell populations from large experiments or clinical samples with the depth and resolution of single-cell RNA sequencing," according to its developers at the University of California, Berkeley.