New program translates grammar of messages sent between cells

March 30, 2021
The messages sent between cells can be translated.(Pixabay/Cassiopeia Arts)

The messages sent between cells can be translated.(Pixabay/Cassiopeia Arts)

Biologists and mathematicians have joined forces to create CellChat, a tool that interprets communications from one cell to another with more nuance than previous translation programs, using social-network theories to untangle complex molecular signals.

In a paper published in February in Nature Communications, scientists at University of California, Irvine, offered CellChat as an open-source solution for understanding the sometimes overwhelming amount of data generated by comprehensive analyses of cells. They said the tool could be used to identify the molecular mechanisms that underlie a wide range of diseases.

One way a cell can communicate with others is by releasing signaling chemicals, which include hormones and neurotransmitters, and reading those from other cells using receptors on its cellular membrane. The molecules may travel only to a cell's neighbors or could travel throughout the body, and the messages they carry inform other cells' behavior.

To listen in on these intercellular interactions, the developers of CellChat took advantage of data from single-cell RNA sequencing, a relatively new technique that can extract detailed information of how individual cells express their genes. Although it is useful for classifying cells and predicting their growth, the technique's large amounts of data makes it difficult to easily analyze their communication behaviors.

"It's hard to fully comprehend this amount of raw data by just looking at it," said study author Maksim Plikus, a professor of developmental and cell biology at UC Irvine. "It's exciting because there's a lot of information, but how does one interpret it?"

CellChat uses the single-cell RNA sequencing data to predict the most important messages this type of cell sends out and what kind of messages it can receive, based on the signaling chemicals the cells express and other information. It employs graph theory, which is often used to analyze structures within a social network, as well as pattern recognition to make its predictions.

Plikus described his program as a Google Translate that interprets different kinds of signaling molecules, or "words," in the context of the "sentences" they are used in to understand what cells are communicating to each other. Notably, CellChat can pick up on chemicals that can modify signals to create different meanings for the cell or alter its importance. It can be applied to essentially any system or tissue with single-cell data, according to the professor.

"CellChat converts molecular language of cells to interpretable language of humans — at least for the bioinformaticians," Plikus said. 

Plikus and his team have made CellChat open-source and created a website that lays out the results of previous CellChat analyses of cell communications.

An important potential use of the eavesdropping program, the authors said, is identifying possible causes of diseases on a molecular level that can be subsequently tested. 

In one case, they gave CellChat data from lesional skin of patients with atopic dermatitis, a form of eczema that causes inflamed skin, as well as nonlesional skin cells for comparison. The program made fresh predictions on which signals were most active in only the lesional skin cells, which could explain how the disease occurs.

"If there are strong signals only present in diseased skin, that means that if you silence this, very likely this could be a novel therapeutic strategy," Plikus said. "So this could be really useful for speeding up therapy discovery."

The researchers compared CellChat's predictive abilities against other programs that infer cell communication, such as SingleCellSignalR, iTALK and CellPhoneDB. The tools were fed datasets from mouse skin cells, and their outputs were compared with the "correct answers" already generated by other approaches.

CellChat was found to have higher accuracy and lower rates of false positives than its peers when predicting cellular communications. It was also better at capturing stronger messages, though it picked up on fewer overall. 

Plikus said his team also wants to improve the tool so it can overcome a blind spot of modern analysis tools, which process cellular signaling factors but do not take into account the relative position of cells. Such an approach produces false positives because it may come to the conclusion that a cell responds to a message despite the communicating cell being too far away for the message to be received.

The study, "Inference and analysis of cell-cell communication using CellChat," published Feb. 17 in Nature Communications, was authored by Suoqin Jin, Christian Guerrero-Juarez, Lihua Zhang, Ivan Chang, Raul Ramos, Maksim Plikus and Qing Nie, University of California, Irvine; Chen-Hsiang Kuan, University of California, Irvine and National Taiwan University; and Peggy Myung, Yale University. 

We use cookies to improve your experience on our site and to show you relevant advertising.