Heavy-tailed kernels reveal a finer cluster structure in t-SNE visualisations

T-distributed stochastic neighbour embedding (t-SNE) is a widely used data visualisation technique. It differs from its predecessor SNE by the low-dimensional similarity kernel: the Gaussian kernel was replaced by the heavy-tailed Cauchy kernel, solving the 'crowding problem' of SNE. Here, we develop an efficient implementation of t-SNE for a t-distribution kernel with an arbitrary degree of freedom ν, with ν → ∞ corresponding to SNE and ν = 1 corresponding to the standard t-SNE. Using theoretical analysis and toy examples, we show that ν < 1 can further reduce the crowding problem and reveal finer cluster structure that is invisible in standard t-SNE. We further demonstrate the striking effect of heavier-tailed kernels on large real-life data sets such as MNIST, single-cell RNA-sequencing data, and the HathiTrust library. We use domain knowledge to confirm that the revealed clusters are meaningful. Overall, we argue that modifying the tail heaviness of the t-SNE kernel can yield additional insight into the cluster structure of the data.

Single-cell connectomic analysis of adult mammalian lungs

Efforts to decipher chronic lung disease and to reconstitute functional lung tissue through regenerative medicine have been hampered by an incomplete understanding of cell-cell interactions governing tissue homeostasis. Because the structure of mammalian lungs is highly conserved at the histologic level, we hypothesized that there are evolutionarily conserved homeostatic mechanisms that keep the fine architecture of the lung in balance. We have leveraged single-cell RNA sequencing techniques to identify conserved patterns of cell-cell cross-talk in adult mammalian lungs, analyzing mouse, rat, pig, and human pulmonary tissues. Specific stereotyped functional roles for each cell type in the distal lung are observed, with alveolar type I cells having a major role in the regulation of tissue homeostasis. This paper provides a systems-level portrait of signaling between alveolar cell populations. These methods may be applicable to other organs, providing a roadmap for identifying key pathways governing pathophysiology and informing regenerative efforts.

Single-Cell Analysis Reveals a Hair Follicle Dermal Niche Molecular Differentiation Trajectory that Begins Prior to Morphogenesis

Delineating molecular and cellular events that precede appendage morphogenesis has been challenging due to the inability to distinguish quantitative molecular differences between cells that lack histological distinction. The hair follicle (HF) dermal condensate (DC) is a cluster of cells critical for HF development and regeneration. Events that presage emergence of this distinctive population are poorly understood. Using unbiased single-cell RNA sequencing and in vivo methods, we infer a sequence of transcriptional states through which DC cells pass that begins prior to HF morphogenesis. Our data indicate that Wnt/β-catenin signaling is required to progress into an intermediate stage that precedes quiescence and differentiation. Further, we provide evidence that quiescent DC cells are recent progeny of selectively proliferating cells present prior to morphogenesis and that are later identified in the peri-DC zone during DC expansion. Together, these findings provide an inferred path of molecular states that lead to DC cell differentiation.