BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:BSU Seminar: “Statistical learning for drug screening in persona
 lized cancer therapy” - Manuela Zucknick\, University of Oslo
DTSTART:20181002T143000Z
DTEND:20181002T153000Z
UID:TALK111526@talks.cam.ac.uk
CONTACT:Alison Quenault
DESCRIPTION:Large-scale cancer pharmacogenomic screening experiments profi
 le hundreds of cancer cell lines versus hundreds of clinically approved or
  experimental compounds to study drug sensitivity and/or synergistic effec
 ts of drug combinations. The aim of these in vitro studies is to use the g
 enomic profiles of the cell lines together with information about the drug
 s to predict the response of individual cell lines to a particular drug or
  combination of drugs\, and ultimately to learn about in vivo treatment re
 sponse for patients.\n\nThis is a multi-task multi-view prediction problem
  where there is only little predictive value in each of the individual dat
 a sets. Therefore\, it is important to optimize prediction performance by 
 combining the different data sources efficiently\, by borrowing informatio
 n across experiments\, and by using external knowledge wherever available.
  A naïve approach to address this problem is to vectorise all available d
 ata and to apply standard methods for high-dimensional linear regression\,
  but vectorisation will easily “blow up” the regression coefficient ve
 ctor and lead to a very inefficient use of the data.\n\nOur task is made e
 asier by the fact that there is strong structure in the data due to the ex
 perimental setup and biological and biochemical constraints\, and an effic
 ient use of this structure can massively reduce the number of effective pa
 rameters that need to be estimated. This is crucial\, since the sample siz
 e is typically much smaller than the number of input features. The data st
 ructure is often assumed to be non-linear and tends to be expressed in mul
 ti-way arrays (or tensors). In this talk I will discuss some approaches th
 at can help us to build regression models which capture this structure eff
 iciently. I will present examples\, where we impose structure either throu
 gh the regression coefficient array by designing structured penalty terms 
 or priors\, or through direct modification of the input data array\, e.g. 
 in multiple kernel learning.
LOCATION:Large Seminar Room\, 1st Floor\, Institute of Public Health\, Uni
 versity Forvie Site\, Robinson Way\, Cambridge
END:VEVENT
END:VCALENDAR