license: other
datasets:
- bostongene/Kassandra-data
Kassandra: Cell Deconvolution Tool from RNA-seq
Precise reconstruction of the tumor microenvironment using bulk RNA-seq and a unique machine learning-based algorithm trained on artificial transcriptomes
Here, we present Kassandra, a robust and accurate cell deconvolution tool developed for analysis of healthy tissue and tumor biopsies. Based on RNA-seq NGS data of a biological sample, Kassandra predicts cellular composition including stromal and immune elements by analyzing the gene expression. This will lead to an improved understanding of the tumor microenvironment, which is a critical factor in cancer pathogenesis, clinical outcome, and therapeutic resistance.
Kassandra is a decision tree machine learning-algorithm trained on a collection of over thousands of RNA profiles from various sorted cell types. Performance was validated on over 4,000 H&E tissue slides and more than 1,000 samples comprising normal and tumor tissues by comparison with cytometric, immunohistochemical or single-cell RNA sequencing measurements of the same tissue.
Data used in model training
- Collection of 9056 bulk RNA-seq samples from 505 datasets of sorted cells, cancer cells and cell lines
- 348 bulk RNA-seq samples of sorted cell populations (including 343 samples of cells from the blood)
Validation data and cell predictions
- 45 bulk RNA-seq of PBMC or PB samples paired with cytometry data
- Collection of 1750 blood bulk RNA-seq samples from 42 datasets from healthy donors
- 24 validation datasets with 1092 RNA-seq samples paired with flow cytometry, hematological analyzer, CyTOF or other cell percentage data
- Tumor tissues from TCGA project predicted by Kassandra and other algorithms (CIBERSORT, CIBERSORTX, quantiseq, fardeep, ABIS, EPIC, MCP-counter, xcell, Scaden (PBMC))
- Tissues from GTEX project predicted by Kassandra
Licencse
BY UTILIZING THE CODE, YOU ARE CONSENTING TO BE AND AGREE TO BE BOUND BY ALL OF THE TERMS OF THIS LIMITED LICENSE, SEE "LICENSE"