NicheNet: New Tool For Decoding Intercellular Communication

February 14, 2020

By Deborah Borfitz 

February 14, 2020 | A new bioinformatics method couples Google’s most famous search result ranking algorithm with a parameter optimization technique powered by machine learning to give researchers a better picture of intercellular communication—including the effects of extracellular signals (called ligands) on the gene expression of other cells. That makes NicheNet different from other computational approaches based on single-cell expression data from linked sender and receiver cells, says Yvan Saeys, group leader on data mining and modeling for biomedicine at the VIB-UGent Center for Inflammation Research in Flanders, Belgium. 

NicheNet can predict those linkages as well as the effects of signaling on the gene expression of receiver cells by exploiting the enormous amount of available knowledge on intercellular signaling and giving higher weight to the more reliable data sources, explains PhD student Robin Browaeys, who developed the new algorithm. Browaeys was the lead author on a recent study (doi: 10.1038/s41592-019-0667-5) published in Nature Methods introducing the NicheNet method and applying it to tumor and immune cell data to infer active ligands and their gene regulatory effects. 

That level of analysis would be impossible to do manually and arrive at insights, notes Saeys, given that NicheNet is looking at the many known interactions between thousands of genes. On some well-known target genes, a biologist might conduct a literature search to make an educated guess on the most important ligand, says Browaeys, but NicheNet would be required for a genome-wide view of the process and to provide statistical evidence on the guess. 

The R statistical package for NicheNet is available for download from the GitHub page of Saeys’ research unit, so researchers can follow instructions on how to perform a basic NicheNet analysis on their own datasets, says Browaeys. Users need only have modest experience with the R programming language to easily use the tool, he adds, and interest in the research community is high judging from the steady number of inquiries and “interesting results” reported by happy users. 

The limiting factor is that the number of well-studied ligands are in the minority, says Saeys, a situation he expects to reverse itself in the coming years. Moving forward, NicheNet might be used to generate novel hypotheses on links between ligands and targets that could then be validated with additional experiments. 

Significant improvements to NicheNet will be possible with expanded knowledge about the transcription factors regulated by the less-studied ligands, adds Browaeys. More data specific to cell type, including epigenomics, would also be welcome additions. 

Saeys and Browaeys say they are particularly keen on seeing NicheNet applied to spatial transcriptomic data to identify the relative location of cells expressing specific genes. Some ligands are membrane-bound and signaling is limited by the proximity of receiver cells, Browaeys says. Others are secreted, and a subset of those can travel far into the tissue or even through the bloodstream to distant places in the body. 

Researchers have only recently started to look at intercellular communication at all, says Saeys, so the technologies being explored are all “brand new.” These include the dynamics of cell-to-cell signaling over time, says Saeys. “For example, in immune response there will be a whole cascade of different interactions switching on and off between different cell types.” 

Getting Granular 

While developing NicheNet, Saeys’ research team has been collaborating closely with the lab of immunologist Martin Guilliams, also in the VIB-UGent Center for Inflammation Research, to validate the methodology. Last fall, in a study published in Immunity   (, they put NicheNet to the test on data from Guilliams’ lab specific to Kupffer cells—macrophages in the liver bloodstream—and it successfully predicted ligand-receptor interactions. 

Guilliams’ lab is also likely to generate spatial data on the distribution of different cell types in the liver, which NicheNet could ingest to better predict how cells are interacting with each other, says Browaeys. Novel spatial transcriptomic technology is making “much more granular” information available, including from which part of an organ data is being extracted. 

In addition to the liver, researchers have been experimenting with NicheNet on data from brain tissues, tumors, and intercellular activity in the gut, Browaeys says. It has also been used for in vitro stem cell research. 

In oncology, he adds, NicheNet could allow better insights into treatment effects by comparing how cells communicate within a tumor environment before and after therapy—and possibly explain why some drugs work while others are futile. 

The popularity of technologies to extract gene expression bioinformation on single cells has enabled NicheNet to present “a whole new layer of information” on how different types of cells talk to one another, says Saeys. Scientists have only recently learned that immune cells play a bigger role than anticipated in the tumor micro-environment. “I think NicheNet will tell us about the other cells types in the tumor that immune cells are communicating with… that could be used for guiding therapy.”