Fast Interpolation-Based RNA Data Visualization Method

Fast Interpolation-Based RNA Data Visualization Method

Researchers from Yale University accelerated t-distributed stochastic neighbor embedding (t-SNE) for visualization of rare cell populations

A team of researchers from Yale University improved the mathematical formula for a bioinformatics data visualization method. The update may help to develop snapshots of single-cell gene expression several times faster and at much higher-resolution. The improved approach is expected to reduce the rendering time of a million-point single-cell RNA-sequencing (scRNA-seq) data set from over three hours down to fifteen minutes, according to the researchers. t-Distributed Stochastic Neighborhood Embedding (t-SNE) is used to represent patterns in RNA sequencing data that is collected at the single cell level—scRNA-seq data—in two dimensions. t-SNE organizes the cells according to the genes they express and is used to identify new cell types and cell states.

However, t-SNE is a slow process, therefore, scRNA-seq dataset is often reduced as a smaller sample is taken from the initial sample before applying t-SNE. This can compromise the results as the process may fail to capture rare cell populations. Another team of Yale University had previously developed the fast multipole method (FMM) — a numerical technique that accelerates the calculation of long-ranged forces in the n-body problem. The current research team recognized that the principles behind the FMM can be used to nonlinear dimensional reduction problems such as t-SNE, The team used the technique to accelerate t-SNE and renamed it as FIt-SNE, or fast interpolation-based t-SNE.

According to Yuval Kluger, senior author and Yale University professor of pathology, the approach can be used to rapidly analyze single cell RNA-sequencing data and to characterize rare cell subpopulations that cannot be detected when the data is subsampled prior to t-SNE. The team used a heatmap-style visualization for its FIt-SNE results. This can help to observe the expression patterns of thousands of genes at the level of single cells simultaneously. According to the researchers, FIt-SNE can accelerate further work in the field of developmental biology along with neuroscience and cancer research, which require single-cell sequencing for mapping the brain and analyzing tumors. The research was published in the journal Nature Methods on February 11, 2019.


You May Also Like