Fabio Parisi, Ph.D.

Postdoctoral fellow

I received my engineering degree and MSc in bioinformatics from Chalmers University in Göteborg, Sweden. I received my doctoral training as a member of the Swiss Institute of Bioinformatics at the Institute of Experimental Cancer Research in Lausanne, Switzerland, where I obtained his PhD from the University of Lausanne in 2008, under the guidance of Dr. Naef. Since 2008 I have been a member of the laboratory of Dr. Kluger. My research focus on modeling and interpretation of high-throughput data. Some of the topics I am particularly interested in are the identification of not-common genetic markers and the study of clonality.

Identification of not-common genetic markers for diseases (or genetic susceptibility): Complex traits diseases and genetic susceptibility can be the result of the interaction of several inherited alleles. However, it is not always clear whether the set of alleles associated to the trait is fixed and predefined across the entire population, or it is rather depending on the genetic background. We apply machine learning technique to databases of SNPs from families with children affected by autism. Previous analyses have failed to identify significant sets of risk loci. In our analyses we combine extended linkage disequilibrium scores with notions from compressed sensing to identify meaningful manifolds on which to project the data. In these manifolds it is possible to build local discriminators to predict the disease status or the genetic susceptibility.

Deconvolution of mixtures of clones: Tumoral cells within tissue samples are often a collection of different clones. Whether clones present at diagnosis may be associated to later occurring metastases, or even relapse after treatment, is an often overlooked question. Current high-throughput analysis techniques, such as SNP- or expression-arrays, are based on pooling of cells from the sample. The mixing of the different clones makes it very difficult if not impossible to identify single clones from high-throughput data. We are interested in developing mathematical models and computational algorithm to deconvolve the clonal mixture in copy number analyses. To this aim, we are also developing novel segmentation approaches that will identify the regions of piecewise constant abnormal signal (e.g. fixed amount of aberrations in the segment). From the results of the segmentation we infer the most probable clones, each described by the aberration status at each locus, and the relative proportion of each clone, according to a biologially motivated optimization procedure. We are currently testing our findings with images from cytogenetics.

Address:
540 First Avenue
Skirball Institute of Biomolecular Medicine, 3rd Fl., Lab. 7
NYU Langone Medical Center
New York, NY 10016, USA
Phone: +1-212-263- 5764
Fax: +1-212-263-5711