Currently, it is quite challenging to analyze single-cell data comprising of numerous cells and samples and to address varieties emerging from batch impacts and different sample preparations. For this purpose, a group of Yale scientists has developed a new deep neural network called SAUCIE to uncover larger samples of activity of individual cells that originate from a multitude of individuals.
The name SAUCIE is an abbreviation for Sparse Autoencoder for Clustering, Imputation, and Embedding. Scientists developed this neural network to uncover crucial cellular differences within individuals as well as broader patterns that tell the story of how the body functions.
For testing, scientists used SAUCIE to analyze 20 million cells from 60 patients. They identified rare Gamma-Delta T cell types that regulate how the body responds to the virus that causes Dengue fever.
First, author Matt Amodio, a graduate student in computer science, said, “With SAUCIE, we were able to find the proverbial needle in the haystack, and 20 million cells is a massive haystack. The method, which can accommodate a larger volume of patient data than other techniques, will enable identifying larger clusters of cellular activity that could shed light based on a host’s pathologies.”
The paper is published in the journal Nature Methods.