BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Outcome-guided multi-view bayesian clustering for integrative omic
  data analysis
DTSTART:20231103T151500
DTEND:20231103T170000
DTSTAMP:20260413T224216Z
UID:13ec200008572511468bc7a14592d76a6919a5c5549d8cd90bf2d7e6
CATEGORIES:Conferences - Seminars
DESCRIPTION:Paul Kirk\, University of Cambridge\nIntroduction: Although th
 e challenges presented by high dimensional data in the context of regressi
 on are well-known and the subject of much current research\, comparatively
  little work has been done on this in the context of clustering. In this s
 etting\, the key challenge is that often only a small subset of the covari
 ates provides a relevant stratification of the population. Identifying rel
 evant strata can be particularly challenging when dealing with high-dimens
 ional datasets\, in which there may be many variables that provide no info
 rmation whatsoever about population structure\, or - perhaps worse - in wh
 ich there may be (potentially large) variable subsets that define irreleva
 nt stratifications. For example\, when dealing with genetic data\, there m
 ay be some genetic variants that allow us to group patients in terms of di
 sease risk\, but others that would provide completely irrelevant stratific
 ations (e.g. which would group patients together on the basis of eye or ha
 ir colour).\nMethods and Results: Bayesian profile regression is an outcom
 e-guided model-based clustering approach that makes use of a response in o
 rder to guide the clustering toward relevant stratifications. Here we show
  how this approach can be extended to the "multiview" setting\, in which d
 ifferent groups of variables ("views") define different stratifications. W
 e present some results in the context of breast cancer subtyping to illust
 rate how the approach can be used to perform integrative clustering of mul
 tiple 'omics datasets.\nConclusions: When there are multiple clustering st
 ructures present in data\, existing (single view) clustering approaches ca
 n fail to recover the most relevant clustering structure\, even when guide
 d by an appropriate response. Moreover\, traditional variable selection ap
 proaches for clustering do not necessarily improve matters\, since they te
 nd to select variables that define the dominant clustering structure\, reg
 ardless of whether or not it is associated with a response of interest. Re
 al molecular datasets can and do possess multiple clustering structures\, 
 and our outcome-guided multi-view model can allow both relevant and irrele
 vant structures to be identified.\nReferences:\nMolitor\, et al. Bayesian 
 profile regression with an application to the National Survey of Children'
 s Health. Biostatistics. 2010. \nKirk\, Pagani\, Richardson. Bayesian out
 come-guided multi-view mixture models with applications in molecular preci
 sion medicine. arXiv 2023\nKeywords: High dimensional data\, Bayesian clus
 tering\, Omics\n 
LOCATION:MA A3 31 https://plan.epfl.ch/?room==MA%20A3%2031
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR