Abstract 3217: An integrated web server for multi-omics data deconvolution

Tissue data deconvolution has been broadly utilized to characterize the variation of cell components and their relative abundance in cancer tumor microenvironment. Numerous deconvolution methods have been developed with different mathematical and biological considerations. Our previous studies have...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Cancer research (Chicago, Ill.) Ill.), 2020-08, Vol.80 (16_Supplement), p.3217-3217
Hauptverfasser: Lu, Xiaoyu, Dang, Pengtao, Tu, Szu-Wei, Chang, Wennan, Wan, Changlin, Zhang, Chi, Cao, Sha
Format: Artikel
Sprache:eng
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Tissue data deconvolution has been broadly utilized to characterize the variation of cell components and their relative abundance in cancer tumor microenvironment. Numerous deconvolution methods have been developed with different mathematical and biological considerations. Our previous studies have derived a rigorous mathematical condition for a less biased deconvolution. We have demonstrated a cell type can be reliably inferred from an omics data only of the cell type is with distinct signature genes, i.e., the cell type specific expression signature of the selected marker genes of all cell types are orthogonal or at least non-positively associated with each other, which is named as strong or weak “identifiable” conditions. Our analysis suggests the identifiability of a certain cell type is mostly determined by the significance of its gene markers, which is data set and tissue microenvironment dependent. However, such data set specificity has been commonly ignored in existing methods. In this study, we developed a comprehensive web server, namely ISMD (an Integrated web Server for Multi-omics data Deconvolution), to recommend a best deconvolution method and optimize the gene markers and cell types in analyzing a specific data set. The server solved to two fundamental challenges in the practical application of deconvolution methods, namely (1) method selection and (2) parameter optimization. Specifically, ISMD utilized a semi-supervised approach for method recommendation and parameter optimization, which first evaluate the species, sample size, and tissue type of a given data set and test the low rank structure of commonly used cell type specific markers for cell type and signature gene selection. ISMD can handle the blood, normal, inflammatory, and cancer tissue, and central nervous and hematopoietic systems of human and mouse. The semi-supervised knowledge transfer encoded in ISMD enables its application to multi-omics data including general transcriptomics, ATAC-seq and DNA-methylation data. The method has been validated by using a large collection of tissue samples. A user friend web-server has been released through https://github.com/xiaoyulu95/ISMD. We anticipate ISMD can help to increase the specificity and rigorousness in the application of tissue deconvolution methods, to bring more insights of the role of different cell types and their interactions and stimulate the discovery of new therapeutic targets in a tumor micro-environment. Citation Forma
ISSN:0008-5472
1538-7445
DOI:10.1158/1538-7445.AM2020-3217