A comprehensive map of rice quantitative trait nucleotides (QTNs) and inferred QTN effects is published in Nature Genetics this week, providing a genomic resource for one of the world's most agronomically important plants. Scientists from Shanghai Normal University built the map using eight genome-wide association study cohorts, then applied population genetics to show that domestication, local adaptation, and heterosis are all associated with QTN allele frequency changes in rice. They also used data from publications on rice genetics, population genomics, and GWAS cohorts to develop a genome navigation platform for QTN pyramiding and breeding route optimization, which they implement in the improvement of a widely cultivated indica rice variety.
Initialization is critical to the preservation of global data structure when using t-distributed stochastic neighbor embedding (t-SNE) and uniform manifold approximation and projection (UMAP) and is also likely responsible for differences seen in a recent comparison between the two algorithms, scientists from University of Tübingen and Yale University write in this week's Nature Biotechnology. While t-SNE is one of the most widely used tools in single-cell transcriptomics and cytometry, UMAP was highlighted in a 2019 Nature Biotechnology paper for its apparent advantage over t-SNE in consistency and global data structure preservation. In their report, the researchers show that this alleged superiority is linked to initialization choices and that there is currently no evidence that the UMAP algorithm has any advantage over t-SNE in terms of preserving global structure. "Once informative initialization is used, the two algorithms appear to preserve the global structure similarly well, and modern implementations of the two algorithms work with similar speed," they write.