Florian Rohart

PhD in Applied Statistics


Quick Summary

Pont_du_gard Currently a Research Fellow at the Institute of Molecular Biosciences (IMB) in the 'Program in Complex Trait Genomics' group that I joined in June 2017, I am also the course co-ordinator of STAT3306/7306 on ``Statistical Analysis of Genetic Data'', which was offered for the first time in Semester 2 2017 (August-December) at the University of Queensland, Australia.

Since my arrival at the University of Queensland (UQ) some day of Automn 2014 (harsh season with its 20-25 degrees and around 8 rainy days a month), I have been working on the development of new statistical approaches to answer (mainly) biological questions.

Do you need a statistician's input for an exciting project you have? I'm always happy to help, either Academia or outside, as a collaborator or as a consultant, Australia or worldwide. Send me an email!

Research interests

In lay terms, my research interests lie in extracting signal from big noisy data. That is mainly what I have been doing since the start of my research life in 2009. Usually that involves:
  • high dimensional linear (mixed) models
  • variable selection
  • multivariate analysis
  • data integration
For more specifics about I have done so far and get a better idea of what interests me, you can have a look at the publications tab, my UQ, researcherID or google scholar profiles. You can also have a look at my resume and [linkedin]. As a bonus, here is my PhD manuscript (Dec 2012) (under the supervision of both Béatrice Laurent-Bonneau and Magali San-Cristobal) cause I know it is hard to come by things like that!

Featured in

You may have seen me in:

``A molecular classification of human mesenchymal stromal cells (MSC)'', which led to the `Rohart test' (not my choice) that predicts whether a cell is a MSC based on a 16-genes signature. The test is available online for researchers to try, go have a try or have a direct look at what the graphical outputs looks like on a time course data where you can see the loss of the MSC function with our test, pretty cool uh?
``mixOmics: an R package for `omics feature selection and multiple data integration''. mixOmics is growing and had around 18K downloads in 2016 and already more than 14K (August 2017). It is designed to be used for data analysis and data integration, there is a website with a lot of information and tutorials, a mailing list for updates and a bitbucket for issues/complaints. I have developed mixOmics from v6.00 to v6.3.1 (included) with the major developments of the MINT and DIABLO modules (added in v6.0.0), as well as a huge gain in computational time from v6.3.0 onwards (especially the tuning functions). I am still involved in mixOmics (at the time of writing) but this is not my main occupation at the moment.

last modified: