E Pluribus Unum: United States of Single Cells

摘要

Single cell genomic techniques promise to yield key insights into the dynamic interplay between gene expression and epigenetic modification. However, the experimental difficulty of performing multiple measurements on the same cell currently limits efforts to combine multiple genomic data sets into a united picture of single cell variation [1, 2]. The current understanding of epigenetic regulation suggests that any large changes in gene expression, such as those that occur during differentiation, are accompanied by epigenetic changes. This means that if cells undergoing a common process are sequenced using multiple genomic techniques, examining any of the genomic quantities should reveal the same underlying biological process. For example, the main difference among cells undergoing differentiation will be the extent of their differentiation progress, whether you look at the gene expression profiles or the chromatin accessibility profiles of the cells. We reasoned that this property of single cell data could be used to infer correspondence between different types of genomic data. To infer single cell correspondences, we use a technique called manifold alignment. Intuitively, manifold alignment constructs a low-dimensional representation (manifold) for each of the observed data types, then projects these representations into a common space (alignment) in which measurements of different types are directly comparable [3, 4]. To the best of our knowledge, manifold alignment has never been used in genomics. However, other application areas recognize the technique as a powerful tool for multimodal data fusion, such as retrieving images based on a text description, and multilingual search without direct translation [4].