Single-cell RNA sequencing (scRNA-seq) has opened new doors in biology, allowing researchers to study gene expression in individual cells rather than averaging data across thousands or millions of them. This detailed view helps scientists uncover rare cell types, track cell development, and understand diseases at a deeper level. But there’s a catch, analyzing the data is incredibly complex and computationally demanding.

One essential step in scRNA-seq analysis is clustering, where cells are grouped based on similar gene expression profiles. These clusters help researchers identify different cell types or states within a sample. However, many clustering methods rely on stochastic, or random, processes. That means if you run the same clustering algorithm multiple times, you might get different results. This inconsistency can be a major headache when trying to draw solid biological conclusions.

To address this challenge, a research team at the Institute for Basic Science, Korea have developed a new tool called scICE, which stands for single-cell Inconsistency Clustering Estimator. scICE is designed to evaluate the consistency of clustering results and help users zero in on the most reliable clusters. It works by scoring the similarity of results across multiple clustering runs, highlighting those that appear consistently. This is crucial for ensuring that the biological insights gained from the data are based on robust findings.

Overview of the scICE method for consistent clustering set determination

a To estimate clustering consistency across different numbers of clusters, binary search is initially used to find the resolution parameter ranges corresponding to each number of clusters. For instance, the resolution parameter range for detecting seven clusters from blood cell data is narrowed down to [~1.38 × 10⁻⁴, ~7.53 × 10⁻⁴] (blue boxes). From this range, eleven resolution parameters are uniformly sampled (orange circles in the sub-graph). b The IC at each sampled resolution parameter is calculated after obtaining 15 cluster labels at each resolution parameter. Subsequently, the Leiden algorithm is applied to each cluster label iteratively until IC reaches one or the iteration number reaches 150. Among the final IC values of eleven sampled resolution parameters, the minimum IC and corresponding resolution parameter is selected. c By bootstrapping 15 labels 100 times at the selected resolution parameter, 100 IC values are computed to calculate the median IC. When the median IC is lower than the threshold of 1.005 (red dashed line), indicating ~0.25% or less clustering inconsistency, the clustering result is determined as consistent. d After assessing the consistency of clustering for all numbers of clusters, the consistent clustering set (red arrow) is determined by excluding the cluster numbers where the median IC is higher than the threshold (shaded high inconsistency region). The box plot features a central line representing the median, a box defining the interquartile range (IQR) from the 25th to the 75th percentile, and whiskers extending to the furthest data points within 1.5 times the IQR from the box edges (Q1 and Q3). Data points beyond these whiskers are plotted individually as outliers. 

What really sets scICE apart is its speed. Existing consensus clustering tools like multiK and chooseR can be powerful, but they require significant computational resources, especially for datasets with thousands or tens of thousands of cells. scICE achieves up to a 30-fold improvement in speed, making it accessible for large-scale datasets that would otherwise be too time-consuming to process using traditional methods.

In their work, the researchers applied scICE to 48 real and simulated scRNA-seq datasets, some of which included more than 10,000 individual cells. The tool consistently identified reliable clusters, reducing the number of groups scientists needed to explore. This focused approach not only saves time but also reduces the risk of chasing misleading or noisy patterns in the data.

The benefits of scICE are especially relevant as single-cell datasets continue to grow in size and complexity. Whether researchers are studying cancer, brain development, or immune system responses, scICE provides a faster and more reliable way to make sense of the data, ultimately bringing us closer to breakthroughs in personalized medicine, regenerative therapies, and beyond.