Wasserstein Wormhole: Scalable Optimal Transport Distance with Transformers.

Optimal transport (OT) and the related Wasserstein metric ( W ) are powerful and ubiquitous tools for comparing distributions. However, computing pairwise Wasserstein distances rapidly becomes intractable as cohort size grows. An attractive alternative would be to find an embedding space in which pairwise Euclidean distances map to OT distances, akin to standard multidimensional scaling (MDS). We present Wasserstein Wormhole, a transformer-based autoencoder that embeds empirical distributions into a latent space wherein Euclidean distances approximate OT distances. Extending MDS theory, we show that our objective function implies a bound on the error incurred when embedding non-Euclidean distances. Empirically, distances between Wormhole embeddings closely match Wasserstein distances, enabling linear time computation of OT distances. Along with an encoder that maps distributions to embeddings, Wasserstein Wormhole includes a decoder that maps embeddings back to distributions, allowing for operations in the embedding space to generalize to OT spaces, such as Wasserstein barycenter estimation and OT interpolation. By lending scalability and interpretability to OT approaches, Wasserstein Wormhole unlocks new avenues for data analysis in the fields of computational geometry and single-cell biology. Software is available at http://wassersteinwormhole.readthedocs.io/en/latest/.

Novel Approach for Pancreas Transcriptomics Reveals the Cellular Landscape in Homeostasis and Acute Pancreatitis.

BACKGROUND & AIMS: Acinar cells produce digestive enzymes that impede transcriptomic characterization of the exocrine pancreas. Thus, single-cell RNA-sequencing studies of the pancreas underrepresent acinar cells relative to histological expectations, and a robust approach to capture pancreatic cell responses in disease states is needed. We sought to innovate a method that overcomes these challenges to accelerate study of the pancreas in health and disease. METHODS: We leverage FixNCut, a single-cell RNA-sequencing approach in which tissue is reversibly fixed with dithiobis(succinimidyl propionate) before dissociation and single-cell preparation. We apply FixNCut to an established mouse model of acute pancreatitis, validate findings using GeoMx whole transcriptome atlas profiling, and integrate our data with prior studies to compare our method in both mouse and human pancreas datasets. RESULTS: FixNCut achieves unprecedented definition of challenging pancreatic cells, including acinar and immune populations in homeostasis and acute pancreatitis, and identifies changes in all major cell types during injury and recovery. We define the acinar transcriptome during homeostasis and acinar-to-ductal metaplasia and establish a unique gene set to measure deviation from normal acinar identity. We characterize pancreatic immune cells, and analysis of T-cell subsets reveals a polarization of the homeostatic pancreas toward type-2 immunity. We report immune responses during acute pancreatitis and recovery, including early neutrophil infiltration, expansion of dendritic cell subsets, and a substantial shift in the transcriptome of macrophages due to both resident macrophage activation and monocyte infiltration. CONCLUSIONS: FixNCut preserves pancreatic transcriptomes to uncover novel cell states during homeostasis and following pancreatitis, establishing a broadly applicable approach and reference atlas for study of pancreas biology and disease.

scKINETICS: inference of regulatory velocity with single-cell transcriptomics data.

MOTIVATION: Transcriptional dynamics are governed by the action of regulatory proteins and are fundamental to systems ranging from normal development to disease. RNA velocity methods for tracking phenotypic dynamics ignore information on the regulatory drivers of gene expression variability through time. RESULTS: We introduce scKINETICS (Key regulatory Interaction NETwork for Inferring Cell Speed), a dynamical model of gene expression change which is fit with the simultaneous learning of per-cell transcriptional velocities and a governing gene regulatory network. Fitting is accomplished through an expectation-maximization approach designed to learn the impact of each regulator on its target genes, leveraging biologically motivated priors from epigenetic data, gene-gene coexpression, and constraints on cells' future states imposed by the phenotypic manifold. Applying this approach to an acute pancreatitis dataset recapitulates a well-studied axis of acinar-to-ductal transdifferentiation whilst proposing novel regulators of this process, including factors with previously appreciated roles in driving pancreatic tumorigenesis. In benchmarking experiments, we show that scKINETICS successfully extends and improves existing velocity approaches to generate interpretable, mechanistic models of gene regulatory dynamics. AVAILABILITY AND IMPLEMENTATION: All python code and an accompanying Jupyter notebook with demonstrations are available at http://github.com/dpeerlab/scKINETICS.

Epigenetic plasticity cooperates with cell-cell interactions to direct pancreatic tumorigenesis.

The response to tumor-initiating inflammatory and genetic insults can vary among morphologically indistinguishable cells, suggesting as yet uncharacterized roles for epigenetic plasticity during early neoplasia. To investigate the origins and impact of such plasticity, we performed single-cell analyses on normal, inflamed, premalignant, and malignant tissues in autochthonous models of pancreatic cancer. We reproducibly identified heterogeneous cell states that are primed for diverse, late-emerging neoplastic fates and linked these to chromatin remodeling at cell-cell communication loci. Using an inference approach, we revealed signaling gene modules and tissue-level cross-talk, including a neoplasia-driving feedback loop between discrete epithelial and immune cell populations that was functionally validated in mice. Our results uncover a neoplasia-specific tissue-remodeling program that may be exploited for pancreatic cancer interception.

Mapping the evolution of T cell states during response and resistance to adoptive cellular therapy.

To elucidate mechanisms by which T cells eliminate leukemia, we study donor lymphocyte infusion (DLI), an established immunotherapy for relapsed leukemia. We model T cell dynamics by integrating longitudinal, multimodal data from 94,517 bone marrow-derived single T cell transcriptomes in addition to chromatin accessibility and single T cell receptor sequencing from patients undergoing DLI. We find that responsive tumors are defined by enrichment of late-differentiated T cells before DLI and rapid, durable expansion of early differentiated T cells after treatment, highly similar to "terminal" and "precursor" exhausted subsets, respectively. Resistance, in contrast, is defined by heterogeneous T cell dysfunction. Surprisingly, early differentiated T cells in responders mainly originate from pre-existing and novel clonotypes recruited to the leukemic microenvironment, rather than the infusion. Our work provides a paradigm for analyzing longitudinal single-cell profiling of scenarios beyond adoptive cell therapy and introduces Symphony, a Bayesian approach to infer regulatory circuitry underlying T cell subsets, with broad relevance to exhaustion antagonists across cancers.

A gene-environment-induced epigenetic program initiates tumorigenesis.

Tissue damage increases the risk of cancer through poorly understood mechanisms1. In mouse models of pancreatic cancer, pancreatitis associated with tissue injury collaborates with activating mutations in the Kras oncogene to markedly accelerate the formation of early neoplastic lesions and, ultimately, adenocarcinoma2,3. Here, by integrating genomics, single-cell chromatin assays and spatiotemporally controlled functional perturbations in autochthonous mouse models, we show that the combination of Kras mutation and tissue damage promotes a unique chromatin state in the pancreatic epithelium that distinguishes neoplastic transformation from normal regeneration and is selected for throughout malignant evolution. This cancer-associated epigenetic state emerges within 48 hours of pancreatic injury, and involves an 'acinar-to-neoplasia' chromatin switch that contributes to the early dysregulation of genes that define human pancreatic cancer. Among the factors that are most rapidly activated after tissue damage in the pre-malignant pancreatic epithelium is the alarmin cytokine interleukin 33, which recapitulates the effects of injury in cooperating with mutant Kras to unleash the epigenetic remodelling program of early neoplasia and neoplastic transformation. Collectively, our study demonstrates how gene-environment interactions can rapidly produce gene-regulatory programs that dictate early neoplastic commitment, and provides a molecular framework for understanding the interplay between genetic and environmental cues in the initiation of cancer.

Recovering Gene Interactions from Single-Cell Data Using Data Diffusion.

Single-cell RNA sequencing technologies suffer from many sources of technical noise, including under-sampling of mRNA molecules, often termed "dropout," which can severely obscure important gene-gene relationships. To address this, we developed MAGIC (Markov affinity-based graph imputation of cells), a method that shares information across similar cells, via data diffusion, to denoise the cell count matrix and fill in missing transcripts. We validate MAGIC on several biological systems and find it effective at recovering gene-gene relationships and additional structures. Applied to the epithilial to mesenchymal transition, MAGIC reveals a phenotypic continuum, with the majority of cells residing in intermediate states that display stem-like signatures, and infers known and previously uncharacterized regulatory interactions, demonstrating that our approach can successfully uncover regulatory relations without perturbations.