Broadly my areas of expertise/interest cover the prediction of the properties of biological systems, the use and development of machine learning / data mining methods applied to bioinformatics or other fields (such as Natural Language Processing). In my daily work I perform different tasks including research, data analysis, software development (as developer and small team leader) and dissemination.
I was until recently a Postdoctoral researcher working on modeling the response of patients involved in clinical trials based on their genetic background. I previously worked on mixing informations describing proteins at a global level to the one describing their core components the amino-acids, in order to predict structural and chemical properties of these proteins.
In terms of machine learning technology, I started by experimenting with Support Vector Machines, taking advantage of the capacity of ad-hoc Mercer kernels to integrate various kinds of information in the learning process. I also do fancy penalized GLMs for the purpose of feature selection and deep learning models for their capacity to deal with both high-level representations of input data and multiple classes in classification context (the probabilistic interpretation of the models is not bad either). On a regular basis I like to test and compare different models and prediction architectures with the help of well made libraries such as Theano, scipy.learn etc...

Research experience

Current situation (starting December 2010): Postdoctoral associate at the UMRS-1147, Director Pierre Laurent Puig
Topic: Prediction of patient's response to therapeutic treatments.

January 2006 March 2010: Research fellow in the Machine Learning and Neural Network group of the university of Florence (Italy), under the supervision of Paolo Frasconi.
Topic: Prediction of the chemical modifications of amino-acids combining local sequence and whole protein features.

May 2008 August 2008: Internship in the Hirst Group of Nottingham University (U.K),under the supervision of Prof. Jonathan Hirst.
Topic: Protein features to improve the predictions of Zinc binding and Phosphorylated residues.

2005-2006: 9 month master internship in the Systems Biology group of Pasteur Institute (Paris, France), under the supervision of Prof. Benno Schwikowski.
Topic: Development of a Cytoscape plugin for the intuitive visualisation of protein-protein networks.

2004-2005: 9 month master internship in the inserm unit 567 of Cochin Institute (Paris, France), under the supervision of Claire Francastel.
Topic: Centromeric transcripts and hematopoietic differentiation.

2003-2004: 6 month master internship in "Institut du fer à moulin" (Paris, France), under the supervision of Christel Lutz.
Topic: Activation of Stat proteins by Alk receptor.


2006-2010: PhD in Computer Science (University of Florence, Italy).
2005-2006: M.Sc. Bioinformatics (Paris 7 University, France).
2004-2005: M.Sc. Cell Physiology (Paris 7 University, France).
2000-2004: B.Sc. Cell Biology and Physiology (Paris 6 University, France).


Computer Science Bioinformatics & Biology Languages & hobbies
Programing: C, C++, Java
Python, Mathematica, Matlab, R, gnuplot
(scripting) bash, perl, python.

Databases: SQL (MySQL, PostgreSQL).
Markup Languages: xml, html.

Data Mining technics.
Neural Networks
Deep Learning
Support Vector Machines
and Kernel-based algorithms.
Sequence analysis, annotation:
GenScan, Artemis, Pfam...
Sequence comparison and retrieval:
Blast, Fasta, HMM models...
Protein visualisation.

Wet-lab protocols:
Nothern-blots, shifts, Western-blots
Facs/analysis, RNA-fish, cell-culture,
molecular biology, si-RNA regulation,
English: fairly fluent
Italian: fairly fluent
German: notions
French: native speaker

Infography, Scientific illustrations

Talks and Publications

Pharmacogenetic study of digestive and hematologic toxicities induced by FOLFOX in colorectal cancer patients included in the PETACC8 trial. (French) Paris, JFHOD conference 2013.

Argumentative insights from an opinion classification task on a French corpus , to be printed as part of the LENLS 2013 proceedings, Tokyo, Japan.

Pharmacogenomics: Genome-wide transcriptomic variations of human lymphoblastoid cell lines: insights from pairwise gene-expression correlations

GWA studies, statistical learning and Group-Lasso. Paris, MAP5 workgroup (March 11 2011)

Ph.D thesis: Machine Learning Applied To The Prediction Of Amino-acids Properties: study of relevant features and design of Ad-hoc Mercer kernels. Thesis Defence in Florence, Italy (March 3 2010)

Improving prediction of phosphoacceptor sites using global protein descriptors. MALIOB, Turin Italy (January 17 2009)

BMC bioinformatics: A simplified approach to disulfide connectivity prediction from protein sequences.

Using domain, Function and Interaction data to enhance local residue property predictions. Mini Euro CBBM, Rome, Italy (August 8 2008)

CK2 phosphoacceptor sites prediction. EURO XXII, Prague, Czech Republic (July 8 2007)