Cell-cell communication (CCC) can reveal diverse aspects of life processes and cellular function. CCC involves information exchange among cells via ligand-receptor (L-R) pairs. Existing CCC studies rely on gathering L-R pairs to construct databases and analyze L-R gene co-expression. However, building comprehensive L-R databases for different species is challenging, especially considering the sensitivity of CCC inference to database completeness and accuracy.
Researchers led by Prof. LIU Xiaoping from Hangzhou Institute for Advanced Study, UCAS have proposed MDIC3 (Matrix decomposition to infer cell-cell communication), an unsupervised tool for investigating cell-cell communication in any species, which provides a new perspective on inferring cell-cell communication and the results are not limited by specific L-R pairs or signaling pathways.
Using matrix decomposition, MDIC3 can separate information about the genes and cells contained in the expression profile matrix and then obtain cellular communication networks among individual cells. The core of MDIC3 lies in determining the regulatory relationships among cells based on the regulatory relationships among genes.
Overview of MDIC3 algorithm.
The regulatory relationships among genes can be depicted by the gene regulatory network. GRN can cover most of the known and unknown gene regulatory information, among which the secretion of L-R and the activity of downstream factors are also included in the results of gene regulation.
Communications between cell types inferred by MDIC3 are based on the global integration of the communication network between individual cells. This derivation in a more detailed dimension greatly enhances the plausibility of the MDIC3 results and provides an effective tool for studying the interactions between individual cells.
The study was published in Patterns.
MDIC3 method does not require any prior biological information, such as the L-R database. Consequently, MDIC3 is not impacted by the quality of existing L-R information. In contrast, methods investigate cell-cell communications based on L-R co-expression making the final results more sensitive to the completeness and accuracy of L-R information.
MDIC3 does not rely on prior information, allowing it to be applied to any species. However, most other methods investigate cell-cell communications based on L-R information, limiting their applicability to species for which L-R information is available.
MDIC3 infers that cell-cell communications do not depend on specific L-R signaling and at the global cell level. MDIC3 is in line with the biological nature of cell-cell communication, as cell-cell communication is the result of interactions between individual cells via multiple L-Rs and pathways.
The researchers?showed the overall capabilities of MDIC3 by applying it to both mouse and human scRNA-seq datasets and evaluated its reliability and effectiveness using a large body of literature.
Case study on the human and mouse datasets.
They applied MDIC3 to analyze the communication networks during the early embryonic development of mice and found that the dominant communication signaling during the developmental stages from E13.5 to E14.5 changes from dermal cells to epidermal cells.
They used MDIC3 for two datasets of the lesional and nonlesional skin of atopic dermatitis patients to compare the communication networks in different disease conditions. Their findings indicate that inflammatory phenomena may be present in both lesional and nonlesional skin, but the communications among inflammatory cells are more active in lesional skin.
Joint learning at different development stages and in different disease conditions.
To better demonstrate the applicability to any species of MDIC3, they applied MDIC3 to zebrafish species and found that the significant cell-cell communications at the 4 hpf zebrafish embryonic stage were centered around epiblast cells, which is consistent with reported biological features.
The communication networks among cell types in 4 hpf zebrafish embryo inferred by MDIC3.
Contact: LIU Xiaoping
xpliu@ucas.ac.cn
