Modern biology research can now generate valuable data on an unprecedented scale. But there is one major obstacle: making sense of these complex high-dimensional datasets. A team at the University of Basel, Switzerland, has developed a new tool that can provide accurate pictures of the structures hiding within highly complex datasets. By helping scientists uncover hidden patterns in the data, the software will help to make new scientific discoveries.

Over the past decade, biology has entered the era of "big data". Today's high-throughput technologies generate vast datasets that capture the identities and multi-dimensional characteristics of cells. Using single-cell RNA sequencing, for example, researchers can measure the activity of tens of thousands of genes in hundreds of thousands or even millions of individual cells. Similar large-scale datasets are now produced in many fields of biology, from genetics and cancer research to neuroscience.

These data promise to provide fundamental new insights into how cells develop, communicate, and change during disease. However, analyzing and interpreting such complex datasets remains a big challenge that researchers continue to struggle with. "People are good at recognizing patterns in two or three dimensions," says Professor Erik van Nimwegen. "But we simply can't make a picture of a dataset that exists in 10,000 dimensions, and lack intuition for what kind of structures can even exist in such high-dimensional spaces."

In Nature Biotechnology, the researchers present their newly developed software tool "Bonsai", which visualizes high-dimensional data on a tree. They demonstrate that this visualization provides a faithful picture of the structure in the data, including how cells are related and how they may have developed from precursor cells.

A tree instead of a flat map

Most popular tools force data with thousands of dimensions into a two-dimensional map. Although these methods are used in virtually every study, researchers appreciate that such pictures invariably distort the data, making it impossible to tell whether the displayed relationships between cells are true or artefacts created by forcing the data into a two-dimensional visualization. Bonsai overcomes this problem.

Instead of creating a flat map, our tool builds a branching tree, with individual cells at the leaves of the branches. Crucially, the distances along the branches accurately reflect how closely cells are related in the high-dimensional space." The team tested the software on both simulated and real single-cell RNA sequencing datasets. Compared to existing methods, Bonsai reconstructs developmental pathways vastly more accurately, preserves the true relationships between cells, and identifies similar cells much more reliably."

Dr. Daan de Groot, first author

From better pictures to new discoveries

When the researchers analyzed human blood cells, Bonsai not only automatically recovered the known relationships between different blood cell types but also uncovered a novel subtype of a type of immune cells, known as natural killer (NK) cells. The molecular signature of this new subtype of NK cells revealed that they originate from a lineage (called the myeloid lineage) that is different from the so-called lymphoid lineage that all NK cells were thought to derive from.

"This example shows why faithful representation and visualization of complex data matters," says van Nimwegen. "When you can trust the picture, you have a much better chance of making new discoveries."

Because Bonsai can be applied to any type of high-dimensional data, including not only gene expression and chromatin state data but also medical data, data from microbiology (i.e. species composition), or data from neuroscience (i.e. neural firing patterns), the researchers believe Bonsai can be a valuable tool for scientists across the life sciences. The Bonsai software is freely available to the research community at: https://bonsai.unibas.ch/

Source:

Journal reference: