Our genes don’t just work like simple on/off switches—they can be finely tuned through a process called alternative splicing, which allows a single gene to produce multiple versions of a protein. This process is crucial for cell function and is especially important in diseases like cancer, where abnormal splicing can lead to aggressive cell behavior.

But studying alternative splicing at the single-cell level has been a major challenge.

Why Is Studying Splicing at the Single-Cell Level Difficult?

Most studies of alternative splicing look at large groups of cells, which can hide important differences between individual cells. Since each cell type can have its own unique splicing patterns, researchers need methods that can analyze splicing at the single-cell RNA sequencing (scRNA-seq) level. However, scRNA-seq data is tricky to work with because it often has low and uneven coverage—meaning that not all parts of a gene are read equally, making it hard to detect alternative splicing events. Recently, researchers at Ghent University have developed ELLIPSIS to overcome these challenges.

What Makes ELLIPSIS Different?

ELLIPSIS is designed to overcome these challenges by:

• Using locally observed read coverage: It makes the most of the data available, even if coverage is uneven.
• Leveraging conservation of flow: This means it accounts for natural patterns of splicing to improve accuracy.
• Ensuring intra-cell type similarity: It recognizes that cells of the same type tend to have similar splicing patterns, helping it detect differences more reliably.
• Identifying novel splicing events: This is especially important in cancer, where abnormal splicing can lead to unique tumor characteristics.

Overview of the different inputs, outputs and analysis steps required for differential splicing detection

Given the reference annotation and the STAR alignment results, ELLIPSIS builds a splice graph for each gene using 3 consecutive steps: building reference graphs, enriching the graphs with de novo splicing events, and simplifying the graphs by removing unobserved exons and junctions. Similar cells are identified with Seurat findNeighbors. Next, ELLIPSIS computes Ψ-values, which can then be analyzed further in post-processing to identify differentially spliced genes.

To test ELLIPSIS, researchers applied it to simulated data and found that it reliably measured splicing differences between cell types. Then, they used it on glioblastoma, a highly aggressive brain cancer, to compare splicing between cancer cells in different parts of the tumor.

Their findings were striking:

• Cells in the tumor core had different splicing patterns compared to cancer cells infiltrating peripheral tissues.
• The genes affected by these splicing differences were linked to cell movement, shape, and interaction with their environment—all of which influence how the cancer spreads.

By improving how we analyze splicing at the single-cell level, ELLIPSIS could help researchers understand how cancer cells adapt and invade new areas. This could eventually lead to better diagnostic tools and even new therapeutic targets.

With tools like ELLIPSIS, scientists are getting a clearer picture of the complexity of cancer—and taking a step closer to more effective treatments.

Availability – ELLIPSIS quantification tool: https://github.com/MarchalLab/ELLIPSIS.git