Dataset Information

Signature-scoring methods developed for bulk samples are not adequate for cancer single-cell RNA sequencing data.

ABSTRACT: Quantifying the activity of gene expression signatures is common in analyses of single-cell RNA sequencing data. Methods originally developed for bulk samples are often used for this purpose without accounting for contextual differences between bulk and single-cell data. More broadly, few attempts have been made to benchmark these methods. Here, we benchmark five such methods, including single sample gene set enrichment analysis (ssGSEA), Gene Set Variation Analysis (GSVA), AUCell, Single Cell Signature Explorer (SCSE), and a new method we developed, Jointly Assessing Signature Mean and Inferring Enrichment (JASMINE). Using cancer as an example, we show cancer cells consistently express more genes than normal cells. This imbalance leads to bias in performance by bulk-sample-based ssGSEA in gold standard tests and down sampling experiments. In contrast, single-cell-based methods are less susceptible. Our results suggest caution should be exercised when using bulk-sample-based methods in single-cell data analyses, and cellular contexts should be taken into consideration when designing benchmarking strategies.

SUBMITTER: Noureen N

PROVIDER: S-EPMC8916770 | biostudies-literature | 2022 Feb

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Signature-scoring methods developed for bulk samples are not adequate for cancer single-cell RNA sequencing data.

Noureen Nighat N Ye Zhenqing Z Chen Yidong Y Wang Xiaojing X Zheng Siyuan S

eLife 20220225

Quantifying the activity of gene expression signatures is common in analyses of single-cell RNA sequencing data. Methods originally developed for bulk samples are often used for this purpose without accounting for contextual differences between bulk and single-cell data. More broadly, few attempts have been made to benchmark these methods. Here, we benchmark five such methods, including single sample gene set enrichment analysis (ssGSEA), Gene Set Variation Analysis (GSVA), AUCell, Single Cell S ...[more]

PMID: 35212622

Similar Datasets

Project description:BackgroundGlobal patterns of immune cell communications in the immune microenvironment of skin cutaneous melanoma (SKCM) haven't been well understood. Here we recognized signaling roles of immune cell populations and main contributive signals. We explored how multiple immune cells and signal paths coordinate with each other and established a prognosis signature based on the key specific biomarkers with cellular communication.MethodsThe single-cell RNA sequencing (scRNA-seq) dataset was downloaded from the Gene Expression Omnibus (GEO) database, in which various immune cells were extracted and re-annotated according to cell markers defined in the original study to identify their specific signs. We computed immune-cell communication networks by calculating the linking number or summarizing the communication probability to visualize the cross-talk tendency in different immune cells. Combining abundant analyses of communication networks and identifications of communication modes, all networks were quantitatively characterized and compared. Based on the bulk RNA sequencing data, we trained specific markers of hub communication cells through integration programs of machine learning to develop new immune-related prognostic combinations.ResultsAn eight-gene monocyte-related signature (MRS) has been built, confirmed as an independent risk factor for disease-specific survival (DSS). MRS has great predictive values in progression free survival (PFS) and possesses better accuracy than traditional clinical variables and molecular features. The low-risk group has better immune functions, infiltrated with more lymphocytes and M1 macrophages, with higher expressions of HLA, immune checkpoints, chemokines and costimulatory molecules. The pathway analysis based on seven databases confirms the biological uniqueness of the two risk groups. Additionally, the regulon activity profiles of 18 transcription factors highlight possible differential regulatory patterns between the two risk groups, suggesting epigenetic event-driven transcriptional networks may be an important distinction. MRS has been identified as a powerful tool to benefit SKCM patients. Moreover, the IFITM3 gene has been identified as the key gene, validated to express highly at the protein level via the immunohistochemical assay in SKCM.ConclusionMRS is accurate and specific in evaluating SKCM patients' clinical outcomes. IFITM3 is a potential biomarker. Moreover, they are promising to improve the prognosis of SKCM patients.

Project description:The incidence rate and mortality rate of colorectal cancer (CRC) ranks third and second globally. Cancer-associated fibroblasts (CAFs) are the major constituent of the stromal cells in the tumor microenvironment (TME) and are closely associated with patients' prognoses. Our study intended to establish a prognostic model for CRC using hallmark genes of CAFs. The expression values of genes and clinicopathological characteristics of patients were enrolled from the cancer genome atlas database as well as the gene expression omnibus database. The single-cell RNA sequencing data were collected and analyzed in the deeply integrated human single-cell omics database and cancer single-cell expression map databases. The ESTIMATE algorithm was applied to access the infiltration levels of immune and stromal cells. The prognostic genes were selected by the Cox regression analysis and the prognostic signature was constructed by the least absolute shrinkage and selection operator algorithm. gene set enrichment analysis was used to explore the enriched gene sets. In this study, based on bulk RNA sequencing and single-cell RNA sequencing data, and we found that more CAFs were infiltrated in the tumor microenvironment and consisted of 3 subtypes. Then we constructed a prognostic signature for CRC using hallmark genes of CAFs and proved that this signature exhibited high values to predict the overall survival of CRC patients in independent training and validating cohorts. Besides, function enrichment analysis revealed that our prognostic model was significantly associated with immune regulation. Further analysis showed that the infiltrated levels of tumor-suppressing immune cells and the expression of higher immune checkpoint genes in CRC tissues were higher in patients with high-risk scores. Furthermore, immunohistochemistry analysis exhibited that these genes in our prognostic signature were markedly upregulated in CRC tissues. We first constructed a signature based on CAFs hallmark genes to predict the survival of CRC patients and further revealed that the tumor-suppressing microenvironment and dysregulated immune checkpoint genes in CRC tissues were partly responsible for the poor prognosis of patients.

Project description:BackgroundCell death caused by neutrophil extracellular traps (NETs) is known as NETosis. Despite the increasing importance of NETosis in cancer diagnosis and treatment, its role in Non-Small-Cell Lung Cancer (NSCLC) remains unclear.MethodsA total of 3298 NSCLC patients from different cohorts were included. The AUCell method was used to compute cells' NETosis scores from single-cell RNA-sequencing data. DEGs in sc-RNA dataset were obtained by the Seurat's "FindAllMarkers" function, and DEGs in bulk-RNA dataset were acquired by the DESeq2 package. ConsensusClusterPlus package was used to group patients into different NETosis subtypes, and the Enet algorithm was used to construct the NETosis-Related Riskscore (NETRS). Enrichment analyses were conducted using the GSVA and ClusterProfiler packages. Six distinct algorithms were utilized to evaluate patients' immune cell infiltration level. Patients' SNV and CNV data were analyzed by maftools and GISTIC2.0, respectively. Drug information was obtained from the GDSC1, and predicted by the Oncopredict package. Patient response to immunotherapy was evaluated by the TIDE algorithm in conjunction with the phs000452 immunotherapy cohort. Six NRGs' differential expression was verified using qRT-PCR and immunohistochemistry.ResultsAmong all cell types, neutrophils had the highest AUCell score. By Intersecting the DEGs between high and low NETosis classes, DEGs between normal and LUAD tissues, and prognostic related genes, 61 prognostic related NRGs were identified. Based on the 61 NRGs, all LUAD patients can be divided into two clusters, showing different prognostic and TME characteristics. Enet regression identified the NETRS composed of 18 NRGs. NETRS significantly associated with LUAD patients' clinical characteristics, and patients at different NETRS groups showed significant differences on prognosis, TME characteristics, immune-related molecules' expression levels, gene mutation frequencies, response to immunotherapy, and drug sensitivity. Besides, NETRS was more powerful than 20 published gene signatures in predicting LUAD patients' survival. Nine independent cohorts confirmed that NETRS is also valuable in predicting the prognosis of all NSCLC patients. Finally, six NRGs' expression was confirmed using three independent datasets, qRT-PCR and immunohistochemistry.ConclusionNETRS can serves as a valuable prognostic indicator for patients with NSCLC, providing insights into the tumor microenvironment and predicting the response to cancer therapy.

Project description:BackgroundRNA-sequencing has become a standard tool for analyzing gene activity in bulk samples and at the single-cell level. By increasing sample sizes and cell counts, this technique can uncover substantial information about cellular transcriptional states. Beyond quantification of gene expression, RNA-seq can be used for detecting variants, including single nucleotide polymorphisms, small insertions/deletions, and larger variants, such as copy number variants. Notably, joint analysis of variants with cellular transcriptional states may provide insights into the impact of mutations, especially for complex and heterogeneous samples. However, this analysis is often challenging due to a prohibitively high number of variants and cells, which are difficult to summarize and visualize. Further, there is a dearth of methods that assess and summarize the association between detected variants and cellular transcriptional states.ResultsHere, we introduce XCVATR (eXpressed Clusters of Variant Alleles in Transcriptome pRofiles), a method that identifies 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 simulations 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.ConclusionsXCVATR is publicly available to download from https://github.com/harmancilab/XCVATR .

Project description:Pancreatic cancer (PC) is known for its high degree of heterogeneity and exceptionally adverse outcome. While disulfidptosis is the most recently identified form of cell death, the predictive and therapeutic value of disulfidptosis-related genes (DRGs) for PC remains unknown. RNA sequencing data with the follow-up information, were retrieved from the TCGA and ICGC databases. Consensus clustering analysis was conducted on patient data using R software. Subsequently, the LASSO regression analysis was conducted to create a prognostic signature for foreseeing the outcome of PC. Differences in relevant pathways, mutational landscape, and tumor immune microenvironment were compared between PC samples with different risk levels. Finally, we experimentally confirmed the impact of DSG3 on the invasion and migration abilities of PC cells. All twenty DRGs were found to be hyperexpressed in PC tissues, and fourteen of them significantly associated with PC survival. Using consensus clustering analysis based on these DRGs, four DRclusters were identified. Additionally, altogether 223 differential genes were evaluated between clusters, indicating potential biological differences between them. Four gene clusters (geneClusters) were recognized according to these genes, and a 10-gene prognostic signature was created. High-risk patients were found to be primarily enriched in signaling pathways related to the cell cycle and p53. Furthermore, the rate of mutations was markedly higher in high-risk patients, besides important variations were present in terms of immune microenvironment and chemotherapy sensitivity among patients with different risk levels. DSG3 could appreciably enhance the invasion and migration of PC cells. This work, based on disulfidoptosis-related genes (DRGs), holds the promise of classifying PC patients and predicting their prognosis, mutational landscape, immune microenvironment, and drug therapy. These insights could boost an improvement in a better comprehension of the role of DRGs in PC as well as provide new opportunities for prognostic prediction and more effective treatment strategies.

Dataset Information

Signature-scoring methods developed for bulk samples are not adequate for cancer single-cell RNA sequencing data.

Publications

Signature-scoring methods developed for bulk samples are not adequate for cancer single-cell RNA sequencing data.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets