I am an applied statistician, interested in mathematical modeling of biological data, typically employing Bayesian hierarchical methods, latent variable approaches, and generating efficient R packages.
I have several available projects for PhD students for the development of novel statistical methods, and respective software implementations, for analyzing omics data.
This kind of data is particularly interesting to model, because it is characterized by two sources of noise: the biological variability, that is of interest, and the measurement noise, which is nuisance; as a consequence, the original biological process is not observed, and is treated as a latent variable. This requires ad hoc Bayesian approaches, and Markov chain Monte Carlo algorithms, to alternately sample from the conditional distributions of the model parameters, and of the latent variables. Multiple samples can be jointly analyzed and embedded within a Bayesian hierarchical framework, which allows for sharing of information across samples. A further challenge is computational, because thousands of genes are simultaneously analyzed; to limit the runtime of our methods, clever coding techniques will be employed, such as parallel coding, and C++ coding within R (via the Rcpp interface). The methods developed will be validated in simulated and real datasets, and ultimately be released as open-source Biocondoctor R packages.
The ultimate goal of these projects is to design statistical software methods. These tools will then enable biologists to accurately infer and study biological process; this kind of knowledge is key, for instance, when studying diseases, or responses to treatment, ultimately improving medical care.
If you want to know more, drop me an email at Simone.Tiberi@unibo.it or write me on Teams.
For more information, see the following link: https://sites.google.com/view/simonetiberi
