Percentage of cells expressing archetype 5 gastrointestinal gene signatures across intestinal, stomach, and gallbladder epithelial cell types from the CELLxGENE reference atlas

Expression and histological evidence of lipid metabolism in peritoneal metastases, showing up-regulation of lipid anabolism and oxidative stress pathways in archetype 2 cancer cells

Metastasis is the leading cause of cancer deaths. To develop strategies for intercepting metastatic progression, a better understanding of how tumor cells adapt to vastly different organ contexts is needed. To investigate this question, a single-cell transcriptomic atlas of primary tumor and diverse metastatic samples (liver, omentum, peritoneum, stomach wall, lymph node, and diaphragm) from a patient with pancreatic ductal adenocarcinoma who underwent rapid autopsy was generated. Using unsupervised archetype analysis, both shared and site-specific gene programs were identified, including lipid metabolism and gastrointestinal programs prevalent in peritoneal and stomach wall lesions, respectively. We developed PICASSO as a probabilistic approach for inferring clonal phylogeny from single-cell and matched whole-exome sequencing data. Comparison of PICASSO-generated clonal structure with phenotypic signatures revealed that pancreatic cancer cells adapted to local environments with minimal contribution from clonal genotype. Our results suggest a paradigm whereby strong environmental effects are imposed on highly plastic cancer cells during metastatic dissemination.

Summary of genomic and sequencing statistics for the single-nucleus RNA-seq samples analyzed in the main rapid autopsy patient

Validation of lipid metabolic rewiring in peritoneal metastases using single-nucleus RNA-seq from an independent rapid-autopsy patient and phylogenetic analysis revealing lipid anabolism up-regulation in peritoneal lesions and archetype 2 clones across distant clades

Analysis of PDAC cells showing KRAS pathway activity, copy number alterations, and mutational profiles across metastatic tumor sites

MOTIVATION:Sequence simulations along phylogenetic trees play an important role in numerous molecular evolution studies such as benchmarking algorithms for ancestral sequence reconstruction, multiple sequence alignment, and phylogeny inference. They are also used in phylogenetic model-selection tasks, including the inference of selective forces. Recently, Approximate Bayesian Computation (ABC)-based approaches have been developed for inferring parameters of complex evolutionary models, which rely on massive generation of simulated data. For all these applications, computationally efficient sequence simulators are essential. RESULTS:In this study, we investigate fast algorithms for simulating sequences along a phylogenetic tree, focusing on accelerating the speed-limiting component of the simulation process: handling insertion and deletion (indel) events. We demonstrate that data structures which efficiently store indel events along a tree can substantially accelerate the simulation process compared to a naive approach. To illustrate the utility of this efficient simulator, we integrated it into an ABC-based algorithm for inferring indel model parameters and applied it to study indel dynamics within Chiroptera. AVAILABILITY AND IMPLEMENTATION:The source code for the different simulation algorithms, alongside the data used, is available at: https://github.com/nimrodSerokTAU/evo-sim. The simulator has also been integrated into SpartaABC, a website for the inference of indel parameters, accessible at: https://spartaabc.tau.ac.il/.

Reconstructing jets, which provide vital insights into the properties and histories of subatomic particles produced in high-energy collisions, is a main problem in data analyses in collider physics. This intricate task deals with estimating the latent structure of a jet (binary tree) and involves parameters such as particle energy, momentum, and types. While Bayesian methods offer a natural approach for handling uncertainty and leveraging prior knowledge, they face significant challenges due to the super-exponential growth of potential jet topologies as the number of observed particles increases. To address this, we introduce a Combinatorial Sequential Monte Carlo approach for inferring jet latent structures. As a second contribution, we leverage the resulting estimator to develop a variational inference algorithm for parameter learning. Building on this, we introduce a variational family using a pseudo-marginal framework for a fully Bayesian treatment of all variables, unifying the generative model with the inference process. We illustrate our method's effectiveness through experiments using data generated with a collider physics generative model, highlighting superior speed and accuracy across a range of tasks.

Learned representations are often invariant to rotational transformations, leaving individual dimensions non-identifiable and interchangeable. We study how Matryoshka Representation Learning (MRL) induces a task-aligned privileged basis distinct from variance-based or regularizer-induced orderings. In the linear setting, we prove that full-prefix MRL recovers the ordered principal directions, and can be computed efficiently using shared statistics. Empirically, we demonstrate that MRL yields consistent per-dimension structure aligned with task signal, where coordinate magnitude reflects informativeness.

Comprehensive gene set enrichment analysis (GSEA) results showing full enrichment scores for all archetype cluster gene programs

Names of samples, whole-exome sequencing EGA file IDs, and snRNA-seq HTAN file IDs

Overview of the archetype analysis framework and its application to single-cell transcriptomes from metastatic organ sites in the rapid autopsy PDAC dataset

Performance benchmarking of PICASSO showing improved phylogenetic accuracy, computational efficiency, reproducibility, and robustness compared to neighbor-joining approaches

Analysis of archetype 5 cells suggesting possible reseeding of the primary pancreas tumor from the stomach metastasis, integrating expression programs, clonal assignments, and copy-number alterations

Differentially expressed and modular genes for each archetype cluster, including upregulated genes (log₂ fold change > 1, p < 0.05) and Hotspot-identified informative genes with significant autocorrelation (FDR < 0.05)

Lipid metabolic rewiring is a prominent feature of peritoneal metastases. A, UMAP embedding, colored by tissue site (left), and sample distribution and composition (right) of all AC2 cells. B, AC2 hotspot analysis, highlighting lipid metabolism and oxidative stress and detoxification modules. C, Digital pathology of H&E-stained primary and peritoneal metastasis tissue, showing the expansion of adipose tissue in the peritoneum. D, Quantification of adipose and fibrotic tissue in sections in C. E, Cancer clone phylogeny, indicating AC2-enriched clones (purple triangles), fractional tumor site composition for each clone (stacked bars), and the proportion of cells in each clone assigned to AC2 (outer circle).

Profile of a cancer ecosystem from a single patient with PDAC. A, Top, maximal diameter of primary and liver tumors, based on CT measurements at the indicated time points from diagnosis (day 0). Black bar marks the period of mFOLFIRINOX treatment. Bottom, levels of CA19-9 tumor marker in blood, based on indicated measurement days. Baseline at diagnosis (day 0) is 13,000 U/mL, and the upper physiologic limit is 37 U/mL (red line). B, Representative CT scans. Primary and liver metastatic tumors are overdrawn with colored ellipses. L, left; R, right; A, anterior; P, posterior. C, Anatomic location of collected biospecimens used to generate matched snRNA-seq, WES, and H&E data. Circle diameter indicates relative tumor size. D, FDL of cancer cell transcriptomes (45,134 nuclei), colored by sample (Materials and Methods). Stomach refers to stomach wall metastasis.

Assessment of archetype robustness and biological significance, showing consistent archetype gene programs across downsampling tests, their association with MYC, lipid metabolism, and gastrointestinal programs, and comparison with clustering analysis

Overview and benchmarking of IntegrateCNV and PICASSO methods for inferring single-cell copy number alterations and reconstructing subclonal phylogenies in metastatic PDAC

Gene set enrichment analysis (GSEA) scores for Leiden cluster gene programs that were significantly enriched at FDR < 0.05