SCRAM
The SCRAM tool is a hybrid cell annotation method that accurately classifies both tumor and non-tumor cells from Patch-seq and scRNA-seq data, while also identifying hybrid cell states. SCRAM uses a three-step orthogonal process to provide detailed transcriptional and RNA-inferred genomic profiles for each cell: (1) cell type transcriptional annotation using machine-learned neural network models (NNMs); (2) SNV profiling using the XCVATR tool; and (3) CNV calling using the CaSpER and NUMBAT tools. SCRAM NNMs are trained on 11 previously published scRNA-seq (totaling ~1 M cells) human tumor and non-tumor datasets, including developing brain, immune, and glioma cell atlases. Each cell type in the training datasets. Using this methodology, individual cells may be annotated as more than one cell type, which permits for the characterization of hybrid cellular states.
Paper link (Curry et al, Cancer Cell, 2024)
GitHub link
COORS
The reactivation of neurodevelopmental programs in cancer highlights parallel biological processes that occur in both normal development and brain tumors. Achieving a deeper understanding of how dysregulated developmental factors play a role in the progression of brain tumors is therefore crucial for identifying potential targets for therapeutic interventions. Single-cell RNA sequencing (scRNA-Seq) provides an opportunity to understand how developmental programs are dysregulated and reinitiated in brain tumors at single-cell resolution. Here, we introduce COORS (Cell of ORigin like CellS), as a computational tool trained on developmental human brain single-cell datasets that enables annotation of "developmental-like" cell states in brain tumor cells. Applying COORS to various brain cancer datasets, including medulloblastoma (MB), glioma, and diffuse midline glioma (DMG), we identified developmental-like cells that represent putative cells of origin in these tumors. Our work adds to our cumulative understanding of brain tumor heterogeneity and helps pave the way for tailored treatment strategies.
Paper link (Wang et al, biorxiv 2024)
GitHub link
CLIPPR
Accurate classification of cancer subgroups is essential for precision medicine, tailoring treatments to individual patients based on their cancer subtypes. In recent years, advances in high-throughput sequencing technologies have enabled the generation of large-scale transcriptomic data from cancer samples. These data have provided opportunities for developing computational methods that can improve cancer subtyping and enable better personalized treatment strategies. To integrate interpretable features from the bulk and single-cell profiling, we developed an algorithm named CLIPPR which combines the top-performing single-cell models, RNA-inferred copy number variation (CNV) signals, and the initial bulk model to create a meta-model.
Paper link (Shetty et al, Journal of Neuro-oncology 2024)
GitHub link
CaSpER
CaSpER is a signal processing approach for identification, visualization, and integrative analysis of focal and large-scale CNV events in multiscale resolution using either bulk or single-cell RNA sequencing data. CaSpER integrates the multiscale smoothing of expression signal and allelic shift signals for CNV calling. The allelic shift signal measures the loss-of-heterozygosity (LOH) which is valuable for CNV identification. CaSpER employs an efficient methodology for the generation of a genome-wide B-allele frequency (BAF) signal profile from the reads and utilizes it for correction of CNVs calls.
Paper link (Serin Harmanci et al, Nature Communications 2021)
GitHub link
XCVATR
XCVATR (eXpressed Clusters of Variant Alleles in Transcriptome pRofiles) is a method that identified variants and detects local enrichment of expressed variants within embedding of samples and cells in single-cell and bulk RNA-seq datasets, XCVATR visualizes local "clumps" of small and large-scale variants and searches for patterns of association between each variant and cellular states, as described by the coordinates of cell embedding, which can be computed independently using any type of distance metrics, such as principal component analysis or t-distributed stochastic neighbor embedding. Through stimulations and analysis of real datasets, we demonstrate that XCVATR can detect enrichment of expressed variants and provide insight into the transcriptional states of cells and samples. We next sequenced 2 new single cell RNA-seq tumor samples and applied XCVATR. XCVATR revealed subtle differences in CNV impact on tumors.
Paper link (Harmanci et al, BMC Genomics 2022)
GitHub Link