Single-cell reference mapping to construct and extend cell-type hierarchies

Abstract

Single-cell genomics is now producing an ever-increasing amount of datasets that, when integrated, could provide large-scale reference atlases of tissue in health as well as disease. Such large-scale atlases not only increase the scale and generalizability of analyses but also enable combining the knowledge generated by individual studies. Specifically, individual studies often differ in terms of cell annotation terminology and depth, with different groups specializing in different cell type compartments, often using distinct terminology. Understanding how these distinct sets of annotations are related and complement each other would mark a major step towards a consensus-based cell type annotation that reflects the latest knowledge in the field. Whereas recent computational techniques, referred to as 'reference mapping' methods, facilitate the usage and expansion of existing reference atlases by mapping new datasets (i.e. queries) onto an atlas; a systematic approach towards harmonizing dataset-specific cell type terminology and annotation depth is still lacking. Here, we present 'treeArches', a framework to automatically build and extend reference atlases while enriching them with an updatable hierarchy of cell type annotations across different datasets. We demonstrate various use cases for treeArches, from automatically resolving relations between reference and query cell types to identifying unseen cell types not present in the reference, such as disease-associated cell states. We envision treeArches enabling data-driven construction of consensus, atlas-level cell type hierarchies, as well as facilitating efficient usage of reference atlases.

Biologically informed deep learning to query gene programs in single-cell atlases

“…are given in equations ( 12) and ( 17), respectively. Theorem 1 allows calculation of the output of the joint proximal operator Prox ηR α + ηR γ (⋅) in equation ( 8) by applying the proximal operators (equations ( 12), (16) or (19)) on each column of the input of the joint operator independently.…”

“…Single-cell reference mapping is growing in popularity 8,[15][16][17][18] to map query datasets by minimal modification of the reference atlas 19 . Existing reference mapping methods embed new query data into a reference latent space by removing technical differences, such as batch effects between the reference and the query, without access to reference data.…”

Tree-structured topic modelling of single-cell gene expression data uncovers hierarchical relationships between immune cell types

“…Post-hoc methods Methods to build tree-structured relationships of single-cell sequencing data are emerging. treeArches [10] uses single-cell data to construct and extend a cell reference atlas and corresponding hierarchical classifier, while cellTree [11] produces tree-structured relationship hierarchies between cells. These methods construct their relevant cellular tree after estimating their latent representations rather than learning the cellular hierarchies simultaneously with their latent representations.…”

Population-level integration of single-cell datasets enables multi-scale analysis across samples

“…Algorithms for efficient use of reference atlases are known as single-cell reference mapping methods [19][20][21], which build upon data integration algorithms to rapidly update an existing reference atlas by integrating a new query dataset. Transferring information from the reference atlas to the query enables efficient cell-type annotation of the query cells, automatic identification of novel (sub) populations [20,22] and unseen disease or treatment affected cells [4,22,23].…”