Afficher la notice abrégée

Multi-block high-dimensional lasso-penalized analysis with imputation of missing data applied to postgenomic data in an Ebola vaccine trial

hal.structure.identifierVaccine Research Institute [Créteil, France] [VRI]
hal.structure.identifierStatistics In System biology and Translational Medicine [SISTM]
dc.contributor.authorLORENZO, Hadrien
hal.structure.identifierStatistics In System biology and Translational Medicine [SISTM]
hal.structure.identifierUniversité de Bordeaux [UB]
dc.contributor.authorTHIÉBAUT, Rodolphe
hal.structure.identifierInstitut de Mathématiques de Bordeaux [IMB]
hal.structure.identifierQuality control and dynamic reliability [CQFD]
hal.structure.identifierEcole Nationale Supérieure de Cognitique [ENSC]
dc.contributor.authorSARACCO, Jérôme
dc.date.accessioned2024-04-04T03:07:33Z
dc.date.available2024-04-04T03:07:33Z
dc.date.conference2018-01-11
dc.identifier.urihttps://oskar-bordeaux.fr/handle/20.500.12278/193456
dc.description.abstractEnSeveral sets of variables can be analyzed simultaneously by canonical correlation in a multi-way analysis. These sets of variables are often high-dimensional and repeated over time. For instance, full-transcriptome measured by RNA-Seq used to be performed in longitudinal studies as well as other measures such as peptides or cells. Hence, canonical correlation analysis has been extended with regularized approaches to deal with several high dimensional data. However, some measurements can be missing for technical reasons and therefore introduce undesired structures due to the huge dimension of the datasets.Our objective is to find an efficient method allowing to impute the missing values taking into account the three-way structure, participant-transcriptome-time, and also the missing path structure.We proposed an EM-like covariance-maximization lasso-penalized high-dimensional completion matrix algorithm to reach that goal.We compared our approach on simulated data-sets with the mean imputation per gene pertime step, the missMDA-imputeMFA algorithm which takes structure into account and the softImpute solution initially designed to solve the Netix competition a high-dimensional problem. We used two criterions: the L2-error between estimated and simulated values and the L2-error between estimated and simulated covariance matrices. The numerical resultsexhibited the superiority of the proposed method in most of the scenarii. We also illustrated our approach on a real data-set from a phase I Ebola vaccine trial measuring RNA-Seq data after vaccination (richtien, cell report 2017) in 20 participants at 4 different times on whole-blood samples, representing 74 sequenced-samples, among which 24 samples were missing because of technological issues.
dc.language.isoen
dc.title.enMulti-block high-dimensional lasso-penalized analysis with imputation of missing data applied to postgenomic data in an Ebola vaccine trial
dc.typeCommunication dans un congrès
dc.subject.halStatistiques [stat]
dc.subject.halSciences du Vivant [q-bio]/Médecine humaine et pathologie
bordeaux.hal.laboratoriesInstitut de Mathématiques de Bordeaux (IMB) - UMR 5251*
bordeaux.institutionUniversité de Bordeaux
bordeaux.institutionBordeaux INP
bordeaux.institutionCNRS
bordeaux.conference.titleAnnual workshop on Statistical Methods for Post Genomic Data - SMPGD 2018
bordeaux.countryFR
bordeaux.conference.cityMontpellier
bordeaux.peerReviewedoui
hal.identifierhal-01664610
hal.version1
hal.invitednon
hal.proceedingsnon
hal.conference.end2018-01-12
hal.popularnon
hal.audienceInternationale
hal.origin.linkhttps://hal.archives-ouvertes.fr//hal-01664610v1
bordeaux.COinSctx_ver=Z39.88-2004&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.au=LORENZO,%20Hadrien&THI%C3%89BAUT,%20Rodolphe&SARACCO,%20J%C3%A9r%C3%B4me&rft.genre=unknown


Fichier(s) constituant ce document

FichiersTailleFormatVue

Il n'y a pas de fichiers associés à ce document.

Ce document figure dans la(les) collection(s) suivante(s)

Afficher la notice abrégée