CancerSEA is the first dedicated database that aims to comprehensively resolve distinct functional states of cancer cells at the single-cell level. It portrays a cancer single-cell functional state atlas, involving 14 functional states (including stemness, invasion, metastasis, proliferation, EMT, angiogenesis, apoptosis, cell cycle, differentiation, DNA damage, DNA repair, hypoxia, inflammation and quiescence) of 41,900 cancer single cells from 25 cancer types. CancerSEA allows users to query which functional states the gene or gene list of interest is related to in different cancers. Furthermore, it provides PCG/lncRNA repertoires frequently associated with functional states at single-cell resolution across all cancer types, in a specific cancer type and in individual cancer single-cell datasets. Finally, CancerSEA provides a user-friendly interface for comprehensively searching, browsing, visualizing and downloading functional state activity profiles, PCGs/lncRNAs expression profiles and 14 functional state signatures.
All single-cell datasets were collected from public resources (SRA, GEO, and ArrayExpress) (Figure 1A). For scRNA-seq datasets with raw sequencing files, an in-house bioinformatics pipeline was adopted for quality control and expression quantification (Figure 1B). For scRNA-seq datasets with only expression matrix, we directly converted the expression values to TPM/CPM values using a custom script. For quality control, we removed non-malignant single cells based on the metadata and a RNAseq-inferred copy number variation (CNV) approach, and filtered cells with low quality. At last, we retained 41,900 cells from 25 cancer types.
Figure 1
We manually collected 14 function state signatures from databases (HCMDB, Cyclebase, StemMapper, etc) and published literatures, and filtered signature genes to ensure the credibility. Then GSVA and Spearman’s Rank Correlations were used to compute the activities of functional states and the correlations between the activities and gene expressions, respectively. The significant gene-state associations were identified with FDR <0.05 and correlation >0.3.
All data resources were stored in CancerSEA, including expression profiles of all cancer single cells before and after QC, activities of 14 functional states in cancer cells, and 14 states signature genes. Users can browse state atlas and query associations between gene/gene list and functional states in CancerSEA.
Users can search gene (PCG and lncRNA) of interest by entering gene name or Ensemble ID and the relationships between the gene and functional states will be shown.
Click the "Search" button, the search result will be returned.
Users can also search a gene signature to see the relationships between the signature and functional state.
Click the "Submit" button, the search results will be returned, which are the same as those when search a gene.
In the 'Home' page and 'Search' page, users can query a functional state for the associated PCGs/lncRNAs at single-cell resolution. Through clicking the ‘state’ hyperlinks embedded in the 'Home' page, the associated PCGs/lncRNAs across all cancers will be returned.
Users can also search a state for associated PCGs/lncRNAs in a specific cancer from the 'Search' page.
Users can browse functional state atlas of all cancer cells and detailed information of all datasets in the 'Browse' page. Dataset information consisted of 'Detailed description', 'Functional state profile', 'Cell distribution', 'Expression patterns of PCGs/lncRNAs' and 'Inferred CNV heatmap'.
Users can download data resources from CancerSEA in the 'Download' page.
If you have any questions or comments regarding to CancerSEA, you may contact us by sending an email to Yun Xiao (xiaoyun@ems.hrbmu.edu.cn).