BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Talks.cam//talks.cam.ac.uk// X-WR-CALNAME:Talks.cam BEGIN:VEVENT SUMMARY:Incorporating biological information into network inference using structured shrinkage - Gwenael Leday (MRC Biostatistics Unit) DTSTART:20160826T094000Z DTEND:20160826T100000Z UID:TALK67069@talks.cam.ac.uk CONTACT:INI IT DESCRIPTION:High-throughput biotechnologies such as microarrays provide th e opportunity to study theinterplay between molecular entities\, which is central to the understanding of disease biology.The statistical descriptio n and analysis of this interplay is naturally carried out with Gaussiangra phical models in which nodes represent molecular variables and edges betwe en them representinteractions. Inferring the edge set is\, however\, a cha llenging task as the number of parametersto estimate easily is much larger than the sample size. A conventional remedy is to regularize orpenalize t he model likelihood. In network models\, this is often done locally in the neighbourhoodof each node. However\, estimation of the many regularizatio n parameters is often di\;cult andcan result in large statistical unce rtainties. We show how to combine local regularization withglobal shrinkag e of the regularization parameters\, via empirical Bayes (EB)\, to borrow strengthbetween nodes and improve inference. Furthermore\, we show how one can use EB so the level ofregularization may di \;er across an arbitra ry number of prede \;ned groups of interactions. Suchauxiliary informat ion is often available in Biology. It is shown that accurate prior informa tion cangreatly improve the reconstruction of the network\, but need not h arm the reconstruction if wrong. LOCATION:Seminar Room 1\, Newton Institute END:VEVENT END:VCALENDAR