%0 Journal Article %T DEEP LEARNING FROM PHYLOGENIES TO UNCOVER THE EPIDEMIOLOGICAL DYNAMICS OF OUTBREAKS %T Apprentissage profond de phylogénies pour révéler la dynamique des flambées épidémiques %+ Bioinformatique évolutive - Evolutionary Bioinformatics %+ Institut de biologie de l'ENS Paris (IBENS) %+ Hub Bioinformatique et Biostatistique - Bioinformatics and Biostatistics HUB %+ Epidémiologie et modélisation de la résistance aux antimicrobiens - Epidemiology and modelling of bacterial escape to antimicrobials (EMAE) %+ Centre de recherche en épidémiologie et santé des populations (CESP) %+ University of Vienna [Vienna] %+ Institut de Systématique, Evolution, Biodiversité (ISYEB ) %+ Institut de Systématique, Evolution, Biodiversité (ISYEB ) %A Voznica, J. %A Zhukova, A. %A Boskova, V. %A Saulnier, E. %A Lemoine, F. %A Moslonka-Lefebvre, M. %A Gascuel, O. %Z J.V. is supported by Ecole Normale Supérieure Paris-Saclay and by ED Frontières de l’Innovation en Recherche et Education, Programme Bettencourt. V.B. would like to thank Swiss National Science Foundation for funding (Early PostDoc mobility grant P2EZP3_184543). O.G. is supported by PRAIRIE (ANR-19-P3IA-0001). %< avec comité de lecture %@ 2041-1723 %J Nature Communications %I Nature Publishing Group %V 13 %N 1 %P 3896 %8 2022-07-06 %D 2022 %R 10.1038/s41467-022-31511-0 %M 35794110 %K Deep learning %K phylodynamics %K molecular epidemiology %K tree simulation and representation %K HIV %Z Life Sciences [q-bio]/Biodiversity/Populations and Evolution [q-bio.PE] %Z Computer Science [cs]/Bioinformatics [q-bio.QM]Journal articles %X Widely applicable, accurate and fast inference methods in phylodynamics are needed to fully profit from the richness of genetic data in uncovering the dynamics of epidemics. Standard methods, including maximum-likelihood and Bayesian approaches, generally rely on complex mathematical formulae and approximations, and do not scale with dataset size. We develop a likelihood-free, simulation-based approach, which combines deep learning with (1) a large set of summary statistics measured on phylogenies or (2) a complete and compact representation of trees, which avoids potential limitations of summary statistics and applies to any phylodynamics model. Our method enables both model selection and estimation of epidemiological parameters from very large phylogenies. We demonstrate its speed and accuracy on simulated data, where it performs better than the state-of-the-art methods. To illustrate its applicability, we assess the dynamics induced by superspreading individuals in an HIV dataset of men-having-sex-with-men in Zurich. Our tool PhyloDeep is available on github.com/evolbioinfo/phylodeep. %G English %2 https://cnrs.hal.science/hal-03861407/document %2 https://cnrs.hal.science/hal-03861407/file/PhyloDeep_NatureCom_VoznicaEtalGascuel_2022.pdf %L hal-03861407 %U https://cnrs.hal.science/hal-03861407 %~ INSERM %~ PASTEUR %~ MNHN %~ EPHE %~ UNIV-AG %~ CNRS %~ APHP %~ GIP-BE %~ ISYEB %~ UVSQ %~ PSL %~ UNIV-PARIS-SACLAY %~ SORBONNE-UNIVERSITE %~ SORBONNE-UNIV %~ SU-SCIENCES %~ TEST-DEV %~ UNIV-PARIS %~ UNIVERSITE-PARIS %~ PNRIA %~ ENS-PSL %~ EPHE-PSL %~ UVSQ-UPSACLAY %~ UNIVERSITE-PARIS-SACLAY %~ SU-TI %~ ANR %~ PRAIRIE-IA %~ GS-SANTE-PUBLIQUE %~ ALLIANCE-SU