Comprehensive evaluation of transcriptome-based cell-type quantification methods for immuno-oncology

Abstract

Motivation The composition and density of immune cells in the tumor microenvironment (TME) profoundly influence tumor progression and success of anti-cancer therapies. Flow cytometry, immunohistochemistry staining or single-cell sequencing are often unavailable such that we rely on computational methods to estimate the immune-cell composition from bulk RNA-sequencing (RNA-seq) data. Various methods have been proposed recently, yet their capabilities and limitations have not been evaluated systematically. A general guideline leading the research community through cell type deconvolution is missing. Results We developed a systematic approach for benchmarking such computational methods and assessed the accuracy of tools at estimating nine different immune- and stromal cells from bulk RNA-seq samples. We used a single-cell RNA-seq dataset of ∼11 000 cells from the TME to simulate bulk samples of known cell type proportions, and validated the results using independent, publicly available gold-standard estimates. This allowed us to analyze and condense the results of more than a hundred thousand predictions to provide an exhaustive evaluation across seven computational methods over nine cell types and ∼1800 samples from five simulated and real-world datasets. We demonstrate that computational deconvolution performs at high accuracy for well-defined cell-type signatures and propose how fuzzy cell-type signatures can be improved. We suggest that future efforts should be dedicated to refining cell population definitions and finding reliable signatures. Availability and implementation A snakemake pipeline to reproduce the benchmark is available at https://github.com/grst/immune_deconvolution_benchmark. An R package allows the community to perform integrated deconvolution using different methods (https://grst.github.io/immunedeconv). Supplementary information Supplementary data are available at Bioinformatics online.

Cell type specific genetic regulation of gene expression across human tissues

“…1A). Estimating the cell type composition of a tissue biospecimen from RNA-seq 40 remains a challenging problem (15) and multiple approaches for inferring cell type proportions have been proposed (16). We performed extensive benchmarking for multiple cell types across several expression datasets (fig.…”

Integrative analysis identifies candidate tumor microenvironment and intracellular signaling pathways that define tumor heterogeneity in NF1

“…Selected latent variables represent specific immune cell types in the tumor microenvironment Given the limited representation of different NF1 tumor types in our genomic variant dataset, we used alternate gene expression metrics to assess the biological underpinnings of the 73 uncharacterized latent variables. We performed tumor immune cell deconvolution analysis [29][30][31] to identify potential immune infiltration signatures present in the individual tumors using CIBERSORT and MCP-counter ( Supplemental Table 4 ). Specifically, CIBERSORT deconvolution indicated the presence of resting mast cells and M2 macrophages in all tumor types ( Figure 5A).…”

“…We used supervised machine learning with random forests [28] to isolate combinations of such latent variables to identify specific molecular signatures that may describe the underlying biology unique to each tumor type. Finally, we integrated this information with sample-matched variant data, immune cell signatures [29][30][31] , and protein activity predictions [32] to provide additional biological context to the most important latent variables. This approach reveals biological patterns that underlie different NF1 nerve sheath tumor types and candidate genes and cellular signatures associated with NF1 tumor heterogeneity.…”

DAISM-DNNXMBD: Highly accurate cell type proportion estimation with in silico data augmentation and deep neural networks

“…The DAISM-DNN method is also highly customizable and can be tailored to estimate the proportions of a large variety of cell types including those which are difficult for existing methods to estimate due to the lack of marker genes or GEP signature matrices. Moreover, it is able to handle complicated tasks on immune cells with overlapping marker or signature genes that are challenging, if not impossible, for existing methods 23,24 . For RNA-seq datasets, we provided an alternative mode (DAISM-RNA) using purified RNAseq data to augment the data with ground truth cell type porportions (Methods).…”

“…T cells, CD8+ T cells, NK cells, monocytes, neutrophils). The fine-grain cell type results of some methods were mapped to coarse-grain cell types according to a hierarchy of cell types as defined in Strum et al 24 .…”