Dimensionality Reduction for Life-Scientists (Boudewijn Lelieveldt)

Speaker: Boudewijn Lelieveldt (LUMC)
Location: O|2 building, Auditorium

The rapid advancements in single-cell and spatial -OMICS measurement technologies are currently revolutionizing life-sciences, and high-dimensional data collections have become nearly ubiquitous across the life-sciences. However, directly interpreting high-dimensional data poses considerable challenges: as a result, the demand for algorithms that create low-dimensional  representations to help explore and interpret such data has grown significantly. Such algorithms reduce to dimensionality of the data to more interpretable and manageable proportions, and their usage has become widespread in life sciences in general.

This seminar aims to provide a basic understanding of technologies for dimensionality reduction, with a focus how these can be utilized in life-sciences research. The seminar consists of 3 parts:

1. Introduction into the concept of high-dimensional data, and the basics of linear dimensionality reduction.
2. The algorithmic basics of two popular non-linear dimensionality techniques (tSNE, UMAP), explained for entry-level life-scientists
3. A critical perspective on several pros and cons inherent to non-linear dimensionality reduction (scalability, faithful data representation, interpretability). Moreover, a set of recommendations of when and how to correctly use dimensionality reduction, but also with examples of misuse of these techniques in the life-sciences.

Please sign-up using the form below.