BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VTIMEZONE
TZID:Europe/Vienna
BEGIN:DAYLIGHT
DTSTART:20190331T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20181028T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260404T015606Z
UID:5c2662ca366e9570821221@ist.ac.at
DTSTART:20190131T100000
DTEND:20190131T110000
DESCRIPTION:Speaker: Matthew Robinson\nhosted by Nick Barton\nAbstract: A m
 ajor challenge across the sciences is to conduct statistical testing for a
 ssociations within large datasets\, where the number of covariates strongl
 y exceeds the number of observations. My group derives novel statistical m
 odels and develops computational algorithms\, to conduct variable selectio
 n and appropriate effect estimation within large-scale data\, all with the
  aim of addressing long-standing questions in the life sciences. This is p
 articularly important for understanding common complex human diseases\, su
 ch as type-2 diabetes\, obesity\, and cardiovascular disease\, where typic
 ally there are many thousands of genetic and environmental risk factors\, 
 each of small effect. Currently our ability to characterise these risk fac
 tors is limiting our ability to respond optimally\, treat and ultimately p
 revent common disease. In this talk\, I will present a flexible hierarchic
 al Bayesian model we have developed that utilises structured graph paralle
 l programming and residual updating\, enabling all effects to be estimated
  conditionally in large-scale data of any type\, with low compute resource
  requirements. Three general extensions of this hierarchical Bayesian mode
 l are then presented\, each of which is applied to address a long-standing
  biological problem. First\, we allow multiple levels of hierarchy\, which
  enables joint estimation of associations between obesity risk and genetic
  and epigenetic variation. We show that this facilitates accurate characte
 rization of disease progression and identification of individuals who are 
 on a path to disease\, as it leads to the identification of biomarkers of 
 large effect. Second\, we further develop this model to provide a full qua
 ntification of the different genomic components and molecular mechanisms t
 hat underlie common disease risk\, which also provides improved disease pr
 ediction. Third\, we present an approach to model time-to-event data with 
 a Weibull distribution that handles sparsity with spike and slab variable 
 selection\, considers left truncation and right censoring\, and utilizes a
 daptive rejection sampling. This enables more accurate discovery and estim
 ation of survival related genomic marker effects and grants novel insight 
 into the genetic architecture of time-to-disease diagnosis. Finally\, I wi
 ll then outline my future interdisciplinary research goals to facilitate a
  range of large-scale analyses across the sciences\, as well as better und
 erstand early-life genetic effects.
LOCATION:Mondi Seminar Room 3\, Central Building\, ISTA
ORGANIZER:tguggenb@ist.ac.at
SUMMARY:Matthew Robinson: Bayesian hierarchical models for large-scale disc
 overy\, estimation and prediction analysis of common complex disease
URL:https://talks-calendar.ista.ac.at/events/1720
END:VEVENT
END:VCALENDAR