BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:&quot\;Breaking non-identifiability using genetic information : an
  application to metabolite data and gene expression&quot\; - Benjamin Frot
 \, University of Oxford
DTSTART:20161108T143000Z
DTEND:20161108T153000Z
UID:TALK67163@talks.cam.ac.uk
CONTACT:Alison Quenault
DESCRIPTION:We consider the problem of learning a conditional Gaussian gra
 phical model in the presence of latent variables. Building on recent advan
 ces in this field\, we suggest a method that decomposes the parameters of 
 a conditional Markov random field into the sum of a sparse and a low-rank 
 matrix. We derive convergence bounds for this estimator and show that it i
 s well-behaved in the high-dimensional regime as well as “sparsistent”
  (i.e. capable of recovering the graph structure). We then show how proxim
 al gradient algorithms and semi-definite programming techniques can be emp
 loyed to fit the model to thousands of variables. Through extensive simula
 tions\, we illustrate the conditions required for identifiability and show
  that there is a wide range of situations in which this model performs sig
 nificantly better than its counterparts\, e.g. it can accommodate more lat
 ent variables. Finally\, the suggested method is applied to two datasets c
 omprising genetic variants and metabolites levels. We show our results are
  well replicated across datasets and assess their biological relevance usi
 ng an external source of validation. We also apply it to a human gene expr
 ession dataset.
LOCATION:Large  Seminar Room\, 1st Floor\, Institute of Public Health\, Un
 iversity Forvie Site\, Robinson Way\, Cambridge
END:VEVENT
END:VCALENDAR