AccueilAccueil>Anita BURGUN

Information Sciences to support
Personalized Medicine

Advances in precision medicine rely on an increasing amount of heterogeneous data, through the extensive use of imaging, genotyping and electronic health records. Learning Health Systems (LHS) have recently emerged as a potential solution to facilitate cross fertilization of care data and research data, in which health information generated from patients is continuously analyzed to improve knowledge that will be transferred to patient care. Such systems require that individual data from various sources, i.e., clinical data, imaging, genetic data, etc, be integrated at the patient level and semantically merged.

Our team is a multidisciplinary group with backgrounds in applied mathematics, biostatistics, bioinformatics, and medical informatics. As a group we have expertise in data driven methods, including statistical approaches, machine learning, deep learning, and phenome wide association studies, as well as in knowledge driven approaches, including semantic models (such as biomedical and gene ontologies), rule-based systems, and multi-level modeling (e.g., ontology-based functional annotation of genes).

We have developed PheWAS approaches using hospital data to extract and analyse new associations between phenotypic traits and genomic information. Studies are hypothesis driven, or hypothesis-free using a PheWAS approach extended to multimodal data. Moreover, we explore the whole phenome-wide landscape of the subpopulations to search for relevant associations. As EHR data come in both structured and unstructured formats, high throughput high content phenotyping using both types of information is essential for creating accurate phenotypic representations of patients. Building on our previous work, we leverage methods of extracting facts about patients from their narrative reports by representing dependencies between facts and not just considering atomic facts. Our approach integrates Gaussian graphical models, which demonstrated their power in terms of exhibiting characteristic patterns of conditional correlations for high dimensional data.

One specific challenging problem is automated extraction and analysis of temporal ordering between events. Phenotyping can be performed with multiple measurements over time, requiring methods that are flexible in their quantification of phenomena, and consider changes in time. Routine data contain artefactual information, and complex facts (e.g., embedding, statements about clinical facts, and treatment sequences). Moreover, some information such as the duration of a disease is inherently uncertain. We collaborate with researchers in computer science in Canada (GRIIS) and in France to propose a model of temporal events linked with a formal medical ontology and combine it with graph modelling.

Our group develops Artificial Intelligence (AI) methods for clinical decision. We demonstrated that ontology-based reasoning classifies atrial fibrillation alerts with results comparable to expert cardiologists. However, a major shortcoming of existing ontologies is their difficulty to integrate probabilistic notions, a research question that was tackled through extensions of the classical description logics. On the other hand, deep convolutional neural networks ware tested recently for the classification of medical images, with a level of competence comparable to medical doctors. A major limitation is that such learning techniques obtain optimal results in presence of “big” volumes of data. We conduct interdisciplinary research on methods to combine knowledge-based techniques and machine learning, to learn from a limited number of heterogeneous cases, with complex phenotypes, and complex underlying biological mechanisms. We investigate solutions based on patient similarity to solve a diagnostic or a therapeutic problem of a new patient by recalling previous cases that exhibited similar characteristics in clinical data warehouses. This approach is developed in the RHU C’IL-LICO project with Imagine Institute.

We are involved in the development of virtual patient modules in cancer treatment in the FLAG–ERA Information Technology: The Future of Cancer Treatment (ITfoC). In ITFoC (and in the future, in DigiTwins) we develop IT solutions for individualized treatment of cancer patients based on the precision medicine paradigm. The approach combines different modeling methodologies using holistic omics data, and existing cancer data from clinical trials and routine care. We also develop a radiomics axis with the radiology department at HEGP, and explore a deep learning approach in text mining with the LIMSI CNRS group.

We have made many methodological contributions in adaptive designs for early phase or small samples in clinical trials. In the last years several publications have stated that the average for the combined success rate at Phase III and submission has fallen to ~50% in recent years. They have shown that bringing two doses forward into phase III testing increased the probability of success and improved the expected net present value. They have stressed out that if the estimation of dose-response curves were better evaluated, phase III confirmatory trials could fail less. Thus, proposing statistical methods to better estimate the dose recommended to further studies is critical. Our research projects focus on how to better estimate the dose-toxicity and dose-efficacy relationships, this is even more complexes in the setting of small samples as in pediatrics and rare diseases. We develop statistical methods for clinical trials and routine data for better understanding how doses, medical decision and observations are linked and what is the best way to estimate them according to covariables and multiple doses administration over time.

 

Team Leader: Anita BURGUN (PU-PH; Univ Paris Descartes)

 

Team Members:

Researchers: Stéphanie ALLASSONNIERE (PU; Univ Paris Descartes), François ANGOULVANT (Cherch Ass; Univ Paris Descartes), Sarah BERDOT (AHU; AP-HP), Patrice DEGOULET (PU-PH em; Univ Paris Descartes), Jean-François ETHIER (Researcher; Sherbrooke Univ), Anne-Sophie JANNOT (MCU-PH; Univ Paris Descartes), Vianney JOUHET (Chercheur invité; CHU Bordeaux), Sandrine KATSAHIAN (PU-PH; Univ Paris Descartes), Joël MENARD (PU-PH em; Univ Paris Descartes), Bastien RANCE  (MCU-PH; Univ Paris Descartes), Brigitte SABATIER (PH; AP-HP), Maxime WACK (AHU; CDD), Sarah ZOHAR (DR; Inserm).

Technical Staff: Abdelali BOUSSADI (IE, AP-HP), Xiaoyi CHEN (IR; CDD)

Young Researchers: Camille AUPIAIS (PhD Student), Redhouane ABDELLAOUI (PhD Student), Antoine BARBIERI (Post-Doc), Angel BENITEZ COLLANTE (PhD Student), Jean-Emmanuel BIBAULT (PhD Student), Sandrine BOULET (PhD Student),  Anais CHARLES-NELSON (PhD Student), Sarah COHEN (PhD Student), William DIGAN (PhD Student), Marie DE ANTONIO (PhD Student), Nicolas GARCELON (PhD Student), Emma GERARD (PhD Student), Christina KHNAISSER (PhD Student),  Andrea LAZZATI (PhD Student), Vincent LOOTEN (PhD Student), Samir MELLIKECHE (PhD Student), Antoine NEURAZ (PhD Student), Adrien OLLIER (PhD Student), Germain PERRIN (PhD Student), Moreno URSINO (Post-Doc).

Administration : Florence BORDU (AI; Inserm)

Contact : Tel:  33 1 44 27 63 93  Fax: 33 1 44 27 64 21 email

 

Publications :

  •  Riviere MK, Yuan Y, Jourdan JH, Dubois F, Zohar S. Phase I/II dose-finding design for molecularly targeted agent: Plateau determination using adaptive randomization. Statistical methods in medical research. 2018; 27(2):466-479. PubMed [journal] PMID: 26988926
  • Petit C, Samson A, Morita S, Ursino M, Guedj J, Jullien V, Comets E, Zohar S. Unified approach for extrapolation and bridging of adult information in early-phase dose-finding paediatric studies. Stat Methods Med Res. 2018 Jun;27(6):1860-1877.
  • Bibault J.E., Giraud P., Durdux C., Taieb J., Berger A., Coriat R., Chaussade S., Dousset B., Nordlinger B., Burgun A.  Deep Learning and Radiomics predict complete response after neo-adjuvant chemoradiation for locally advanced rectal cancer. Sci Rep. 2018 Aug 22;8(1):12611. doi: 10.1038/s41598-018-30657-6. PMID: 30135549
  • Koval I, Schiratti JB, Routier A, Bacci M, Colliot O, Allassonnière S, Durrleman S. Spatiotemporal Propagation of the Cortical Atrophy: Population and Individual Patterns. Front Neurol. 2018 May 4;9:235. doi: 10.3389/fneur.2018.00235. eCollection 2018. PubMed PMID: 29780348; PubMed CentralPMCID: PMC5945895.
  • Chen X., Faviez C., Schuck S., Lillo-Le-Louet A., Texier N., Dahamna B., Huot C., Foulquie P., Pereira S., Leroux V., Karapetiantz P., Guenegou-Arnoux A., Katsahian S., Bousquet C., Burgun A., Mining patients’ narratives in social media for pharmacovigilance: adverse effects and misuse of methylphenidate.  Front Pharmacol. 2018 May 24;9:541. doi: 10.3389/fphar.2018.00541. eCollection 2018. PMID: 29881351
  • Jannot A.S., Burgun A., Thervet E., Pallet N. The Diagnosis-Wide Landscape of Hospital-Acquired AKI. Clin J Am Soc Nephrol. 2017 May 11. pii: CJN.10981016. doi: 10.2215/CJN.10981016. PMID: 28495862
  • Garcelon N, Neuraz A, Benoit V, Salomon R, Kracker S, Suarez F, Bahi-Buisson N, Hadj-Rabia S, Fischer A, Munnich A, Burgun A. Finding patients using similarity measures in a rare diseases-oriented clinical data warehouse: Dr.Warehouse and the needle in the needle stack. J Biomed Inform. 2017 Jul 25. pii: S1532-0464(17)30176-4. doi: 10.1016/j.jbi.2017.07.016. PubMed PMID: 28754522.
  • Thall PF, Ursino M, Baudouin V, Alberti C, Zohar S. Bayesian treatment comparison using parametric mixture priors computed from elicited histograms. Statistical methods in medical research. 2017; :962280217726803. NIHMSID: NIHMS897070 PubMed [journal] PMID: 28870123, PMCID: PMC5658278
  • Rance B, Canuel V, Contournis H, Laurent-Puig P, Burgun A. Integrating heterogeneous data for cancer research: the CARPEM infrastructure. Appl Clin Inform. 2016 May 4;7(2):260-74. doi: 10.4338/ACI-2015-09-RA-0125. eCollection 2016.
  • Teramukai S, Daimon T, Zohar S. An extension of Bayesian predictive sample size selection designs for monitoring efficacy and safety. Stat Med. 2015 Sep 30;34(22):3029-39.

View all publications

Les Outils