Accurately identifying different types of cells within tissues is essential for understanding how organs function, how diseases progress, and how treatments might work. With the rise of single-cell and spatial omics technologies, researchers can now study individual cells in great detail. However, a major challenge remains: how do we correctly label or annotate these cells based on the complex data they produce?

Many existing methods rely on either reference datasets from single-cell RNA sequencing (scRNA-seq) or a list of marker genes that are known to be associated with certain cell types. While these approaches can be helpful, they each have limitations. Reference datasets may be incomplete or biased, and marker sets don’t always apply across different experiments or tissue types. When used alone, each method increases the risk of mislabeling, which can compromise results.

A new tool called ScInfeR, developed by researchers at the Indian Institute of Technology (IIT) offers a more reliable solution. ScInfeR stands out by combining both types of information—scRNA-seq references and marker gene sets—using a graph-based framework modeled on the logic of neural networks. This approach allows it to “communicate” across layers of data, mimicking how cells interact in biological systems.

Overview of the ScInfeR workflow

(a) ScInfeR framework takes scRNA-seq, or single-cell ATAC-sequencing (scATAC-seq), or spatial omics expression matrix as input for cell type inference; (b) a cell marker set or scRNA-seq reference matrix is used as secondary input for cell type guidance. Both of these can be retrieved from our ScInfeRDB database. In case, scRNA-seq reference is input, ScInfeR could calculate the cell marker set; (c) ScInfeR annotates cells in three steps: building a similarity matrix, assigning cluster-level labels based on marker correlations, and refining annotations at the single-cell level using neighbourhood-weighted expression.

ScInfeR’s smart design enables it to accurately identify both broad and subtle cell subtypes across a variety of omics data, including scRNA-seq, scATAC-seq (which maps open DNA regions), and spatial transcriptomics (which adds information about a cell’s physical location in tissue). It even supports weighted input, meaning researchers can tell the system which markers are more important or reliable.

The tool has been extensively benchmarked against 10 existing methods across more than 100 different cell-type prediction tasks. ScInfeR consistently came out on top, especially when dealing with difficult data types or batch effects, which are common problems in omics studies.

To make ScInfeR even more accessible, the researchers created ScInfeRDB, an interactive, manually curated database that includes high-quality reference data for 329 cell types, covering nearly 2,500 gene markers from 28 different human and plant tissue types. This feature makes it easier than ever for researchers to apply ScInfeR to their own datasets without having to build a custom reference from scratch.

Availability – ScInfeR is available as an R package, with both the tool and database publicly accessible at https://www.swainasish.in/scinfer.