Hadrien Lorenzo

PhD in Biostatistics and engineer in applied Mathematics

Bordeaux University

Biography

I am a PhD in Biostatistics and currently in PostDoc in the SISTM team. I work on analysing high dimensional datasets (just a few individuals and thousands of variables) with longitudinal and or multi-block structures. SVD-based methods are what I love the most and I currently work on dealing with missing values in supervised context.

SISTM is mainly devoted to vaccine research and clinical trial data analyses in the context of largely scaled datasets. This is on of the few INSERM/INRIA teams.

I unreasonably work on the ddsPLS (CRAN-R-package) for which I build a R-Vignette. That vignette is updated as often as I can. This is a titanous task and I also work on a py_ddspls (GitHub-Python-package) which is under developpment. My wish is to be useful to both of the communities which have so much to share!

Interests

Machine Learning
Missing Value Treatment
Biostatistics and genetics
Teaching
Climbing and Mountaineering

Education

PhD in Biostatistics, 2019

Université de Bordeaux
Research Master, Applied Mathematics for Image and Signal Processing, 2015

CentraleSupelec and Paris-Sud University

Projects

Outliers and SIR models

How to detect outliers in SIR models.

Supervised Learning for Multi-Block Incomplete Data

My objective is to deal with entire missing rows in supervised block-structured data sets. As a spicy detail data sets are build on a few participants and a lot of variables.

Thesis memory writing

The objective was to write a thesis memory.

Publications

ddsPLS: A Package to Deal with Multiblock Supervised Problems with Missing Samples in High Dimension

The ddsPLS method considers regression and classification problems in the context of multiblock structured covariate data sets taking …

H Lorenzo, J Saracco, R Thiebaut

PDF

An Original Methodology for the Selection of Biomarkers of Tenderness in Five Different Muscles

For several years, studies conducted for discovering tenderness biomarkers have proposeda list of 20 candidates. The aim of the present …

MP Ellies-Oury, H Lorenzo, C Denoyelle, C Denoyelle, J Saracco, B Picard

PDF Source Document

Supervised Learning for Multi-Block Incomplete Data

In the supervised high dimensional settings with a large number of variables and a low number of individuals, one objective is to …

H Lorenzo, J Saracco, R Thiebaut

PDF Source Document

High-dimensional multi-block analysis of factors associated with thrombin generation potential

The identification of novel biological factors associated with thrombin generation, a key biomarker of the coagulation process, remains …

H Lorenzo, M Razzaq, J Odeberg, PE Moreange, J Saracco, DA Tregouet, R Thiebaut

PDF Source Document

Une PLS multivoie parcimonieuse avec observations manquantes pour données hétérogènes

Several sets of variables can be analyzed simultaneously by canonical correlation in a multi-way analysis. These sets of variables are …

H Lorenzo, R Thiebaut, J Saracco

PDF

Systems Vaccinology Identifies an Early Innate Immune Signature as a Correlate of Antibody Responses to the Ebola Vaccine rVSV-ZEBOV

Predicting vaccine efficacy remains a challenge. We used a systems vaccinology approach to identify early innate immune correlates of …

A Rechtien, L Richert, H Lorenzo, G Martrus, B Hejblum, C Dahlke, R Kasonta, M Zinser, H Stubbe, U Matschl, A Lohse, V Krähling, M Eickmann, S Becker, VEBCON Consortium, R Thiebaut, M Altfeld, M Addo

PDF Source Document DOI

Talks and posters

High-dimensional multi-block analysis of factors associated with thrombin generation potential

The identification of novel biological factors associated with thrombin generation, a key biomarker of the coagulation process, remains …

Jun 6, 2019 12:00 AM 32nd IEEE CBMS International Symposium on Computer-Based Medical Systems – CBMS2019

H Lorenzo, M Razzaq, J Odeberg, PE Morange, J Saracco, DA Trégouët, R Thiébaut

PDF

Apprentissage supervisé pour données massives multi-blocs incomplètes

Les récentes innovations techniques ont permis la production de données massives en biologie, comme les données omiques par exemple …

May 16, 2019 12:00 AM Journée de l’Ecole doctorale EDSP2 2019 - JEDSP2 2019

H Lorenzo, R Thiebaut, J Saracco

PDF

A mono/multi-block sparse PLS for heterogeneous data with missing samples

This is the last update of the current project on dealing with missing samples in multi-block context for supervised datasets.

Jul 5, 2018 12:00 AM Séminaire de Biostatistiques

H Lorenzo, R Thiebaut, J Saracco

PDF

Une PLS multivoie parcimonieuse avec observations manquantes pour données hétérogènes

Several sets of variables can be analyzed simultaneously by canonical correlation in a multi-way analysis. These sets …

Jun 1, 2018 12:00 AM 50èmes Journées de Statistique - JDS 2018

H Lorenzo, R Thiebaut, J Saracco

PDF

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

Several sets of variables can be analyzed simultaneously by canonical correlation in a multi-way analysis. These sets …

Jan 12, 2018 12:00 AM Annual workshop on Statistical Methods for Post Genomic Data - SMPGD 2018

H Lorenzo, R Thiebaut, J Saracco

PDF

Viral Replicative Capacity Displays a Non-Linear Association to Cytokine Profiles Observed in Acutely Infected Volunteers in Contrast to Viral Load and CD4 T-cell Counts

Using the Meso Scale Discovery (MSD) V-PLEX 40 biomarker cytokine multiplex kit, the cytokine profiles of 108 HIV+ Zambian volunteers …

Oct 17, 2016 12:00 AM International AIDS Vaccine Initiative 2016

J Hare, C Streatfield, J Yates, W Kilembe, S Lakhi, H Lorenzo, R Thiebaut, E Hunter, J Gilmour

PDF

Teaching

Genetics project in the context of high dimensional data

That course shows main goals of supervised and unsupervised methods to help students to prepare a project dealing with the 10X data set.

Supervised and unsupervised analysis for high dimensinal data

That course shows main goals of supervised and unsupervised methods and their applications through the spectrum of my proper applications.

PLS and sparse PLS

Introduction to PLS (Partial Least Square) and one of its sparse versions.

Maths and function analysis

Mathematical courses designed to fit to French second year medical students of the Double Cursus Sante Sciences expecting strong mathematical bases.

Maths and linear algebra

Mathematical courses designed to fit to French second year medical students of the Double Cursus Sante Sciences expecting strong mathematical bases.

Statistics and Probability

Mathematical courses designed to fit to second year medical students expecting strong mathematical bases.

Single-cell datasets visualizations : PCA Versus tSNE

Among new RNA sequencing methods, one permits to detail the RNA (or DNA) of a sample at a cellular level. That is the Single-Cell RNA-seq. That course explores methods to visualize those datasets.

Experience

October 2016 – Present

Bordeaux

PhD

SISTM team, U1219

Data analysis and eigen-space decomposition for supervised problems. Compressing many thoughts :

Missing Value + High-dimension = Act carefully

October 2015 – October 2016

Bordeaux

Research Engineer

SISTM team, U1219

Discover public health and its tools.

February 2015 – August 2015

Châtenay-Malabry/Montréal

Bayesian approach selecting optimal classes of structural models in earthquake analysis for beam structures, internship

CentraleSupelec and Montreal Polytechnique University

Finite elements modelisation and real application to bridge movements with Fourier analysis. Great teams and amazing moments!

September 2014 – January 2015

Gif-sur-Yvette

Identification of prognostic factors in hepato-cellular carcinoma surgery, project

CentraleSupelec and GE Healthcare

A team project dedicated to image segmentation using SVM tools and morphological transformation. A work done in cooperation with GE Healthcare and supervised by Arthur Tenenhaus and Laurent Le Brusquet.

June 2014 – September 2015

Nice

Astrophysics - Image Processing for large structure studies, internship

Lagrange Laboratory

Extra large images implying special dedicated tools such as spherical wavelets but also deeply open minds. What else to say ? It is hard!

Skills

R

80%

Statistics

100%

Overleaf

70%

Python

60%

Contact

+335 57 57 13 93
Université de Bordeaux, 146 Rue Léo Saignat, 33000 Bordeaux, France