Meet TTMap, the new tool for gene-expression analysis

In the wake of mapping the entire human genome and the advent of increasingly affordable high throughput technologies, we can now study the expression of tens of thousands of genes simultaneously. This poses new challenges to data analysis. When comparing two groups like healthy individuals and patients, traditional statistical approaches used to compare the samples often miss important genes unless hundreds of samples are analyzed which is difficult and time consuming.

Newer tools use software that runs sophisticated algorithms to scour massive amounts of gene expression data, analyze them, and identify biologically meaningful subgroups of samples. However, these so-called “clustering algorithms” depend on parameters that the user may find difficult to choose, such as the number of expected subgroups, and that introduce bias. They lack stability, and do not perform well on small sample sizes either.

Enter TTMap (“Two-Tier Mapper”), a new algorithm for enhanced analysis of global gene expression datasets. TTMap is particularly suited for small samples sizes and outperforms current methods of clustering analysis in terms of sensitivity and stability – meaning that its pattern detection is both high and consistent.

TTMap was developed by Rachel Jeitziner, the first Mathematics MSc to obtain a PhD in Life Sciences (EDMS), co-supervised by EPFL Professors Cathrin Brisken (ISREC) and Kathryn Hess (Brain Mind Institute).

To develop TTMap, Jeitziner collaborated with cancer biologists and bioinformaticians, and exploited tools from the field of topology, which studies the properties of space that are preserved through deformations like stretching, twisting, crumpling, and bending. Topological data analysis is now used to reduce noise and find patterns in large and complex datasets such as voting preferences and interactions between basketball players.

“It is important to foster interactions between disciplines, to meet the increasingly complex challenges of modern society,” says Cathrin Brisken. “Rachel mastered the languages of several different disciplines during her PhD”.

“TTMap can be readily used to analyze complex, highly variable biological samples, and is very promising for applications in personalized medicine, as one gets an in-depth appreciation of the differences between individual samples,” says Jeitziner.

Professor Brisken’s lab is part of the Swiss Institute for Experimental Cancer Research (ISREC) within the School of Life Sciences at EPFL. ISREC is deeply involved in the Swiss Cancer Center Léman (SCCL), a cancer research consortium that includes the University hospital of Lausanne (CHUV), the Geneva University Hospitals (HUG), the universities of Lausanne (UNIL) and Geneva (UNIGE), and EPFL.

Other contributors

• Inria Saclay
• University of Geneva