Geneformer is a foundation transformer model pretrained on a large-scale corpus of ~30 million single cell transcriptomes to enable context-aware predictions in settings with limited data in network biology. Here, we will demonstrate a basic workflow to work with ***Geneformer*** models. These notebooks include the instruction to: 1. Prepare input datasets 2. Finetune Geneformer model to perform specific task 3. Using finetuning models for cell classification and gene classification application

Doublets are a characteristic error source in droplet-based single-cell sequencing data where two cells are encapsulated in the same oil emulsion and are tagged with the same cell barcode. Across type doublets manifest as fictitious phenotypes that can be incorrectly interpreted as novel cell types. DoubletDetection present a novel, fast, unsupervised classifier to detect across-type doublets in single-cell RNA-sequencing data that operates on a count matrix and imposes no experimental constraints. This classifier leverages the creation of in silico synthetic doublets to determine which cells in the input count matrix have gene expression that is best explained by the combination of distinct cell types in the matrix. In this notebook, we will illustrate an example workflow for detecting doublets in single-cell RNA-seq count matrices.

Many spatially resolved transcriptomic technologies do not have single-cell resolution but measure the average gene expression for each spot from a mixture of cells of potentially heterogeneous cell types. Here, we introduce a deconvolution method, conditional autoregressive-based deconvolution (CARD), that combines cell-type-specific expression information from single-cell RNA sequencing (scRNA-seq) with correlation in cell-type composition across tissue locations. Modeling spatial correlation allows us to borrow the cell-type composition information across locations, improving accuracy of deconvolution even with a mismatched scRNA-seq reference. **CARD** can also impute cell-type compositions and gene expression levels at unmeasured tissue locations to enable the construction of a refined spatial tissue map with a resolution arbitrarily higher than that measured in the original study and can perform deconvolution without an scRNA-seq reference. Applications to four datasets, including a pancreatic cancer dataset, identified multiple cell types and molecular markers with distinct spatial localization that define the progression, heterogeneity and compartmentalization of pancreatic cancer.