Software Design

​

• 2013 - 2017 SYLVESTRA
  Supervision of software development, SAMOGWAS ANR project (PhD student Duc-Than Phan (october 2013 – june 2015) ; junior engineer Jules Teulade (march 2015 – april 2016); service provider (november
  2016 – july 2017)). This software program allows to model the statistical dependences existing within highly correlated massive data. In the case of genetical data, the software program allows to conduct advanced analysis to identify genotype-phenotype relations. It is possible to choose the clustering method  used by the scalable learning algorithm.
  https://www.ls2n.fr/listelogicielsequipe/DUKe/132/
  number of code lines: 10471
   

• 2013 - 2017 SYLVESTRA++
  Supervision of software development, SAMOGWAS ANR project (junior engineer Jules Teulade (march 2015 – april 2016); service provider (december 2016 – july 2017)).
  This software program is an advanced version of the above SYLVESTRA program. Three modalities are provided to the user:
  (i) "manual": choice of the clustering method and parameter tuning for the clustering method both made by the user
  (ii) ”semi-automatic”: choice of the clustering method made by the user and automatic parameter setting for the clustering method
  (iii) "full automatic": consensus-based clustering method based on the three clustering methods currently implemented, automatic parameter settings
  https://www.ls2n.fr/listelogicielsequipe/DUKe/134/
  number of code lines: 10673
   

• 2014 - 2017 SMMB
  Supervision of software development (PhD doctoral works of Clément Niel (october 2014 – september 2017). This software program deals with feature subset selection, in the context of high-dimensional data. The problem is stated as Markov blanket learning. Multiple Markov blankets are learned from subsets of the initial dataset.  A Markov blanket consensus is learned in the final step.
  https://ls2n.fr/listelogicielsequipe/DUKe/128/
  number of code lines: 3089
   

• 2014 - 2017 SMMB-ACO
  Supervision of software development (PhD doctoral works of Clément Niel (october 2014 – september 2017).
  This software program is an advanced version of the above SMMB program. It relies on a combinatorial  optimization technique related to ant colony optimization.
  https://ls2n.fr/listelogicielsequipe/DUKe/1 30 /
  number of code lines: 3256
   

• 2008 - 2011  CFHLC
  Supervision of the development of four successive versions of CFHLC.
  CFHLC is devoted to  learning a specific kind of probabilistic graphical models developed at the Lina, forests of latent tree models. The context is very high dimensional data. The sequential version is scalable. We also maintain parallelized versions. One of the possible applications is the modeling of linkage disequilibrium (i.e. complex dependences) in genetical data.
  Contributors : Raphaël Mourad (versions 1 and 2), Duc-Thanh Phan (version 3), Jules Teulade-Denantes (version 4).
   

• 2010  GWAS-AS
  Supervision of the development of the software suite GWAS-AS.
  GWAS-AS chains various scripts, to automate GWASs.
  Contributor : Thomas Morissseau.
   

• 2003 - 2006   BacTrans²
  Contribution to the development and supervision of the development of the sofware platform BacTrans².
  BacTrans²  is dedicated to the identification and characterization of putative strong promoters in prokaryotic genomes.
  
  BacTrans² is the result of my long-term collaboration with various partners. The main contributor is  Sylvain Demey (april 2005 - september 2006).
  
  To recapitulate, between january 2003 and april 2003, I wrote the very core of the software suite (putative strong promoter extraction in Thermotoga maritima genome). Then, between june 2003 and august 2003, the first version for the visualization component of mRNA secondary structures associated with putative strong promoters was implemented by Guillaume Blin, PhD student at the LINA. From january 2004 to june 2004, Jeanne Cambefort implemented a statistical component devoted to genes harbouring putative strong promoters in Thermotoga maritima. She fulfilled her contribution as a student of the Research Master "Informatique et Génomique" at the Rennes 1 University. Improving the visualization browser cited above together with adding various functionalities related to the client-server architecture were entrusted to Gaël Herry et Arnaud Drazek, both students of the Computer Science Master at the University of Nantes. Then, from september 2004 to december 2004, I had the platform evolve into a generic one. As BacTrans² project aroused the interest from both the bioinformatician and biologist communities, whealthy times finally came. The project benefitted from a fund allocated by the "Pays de la Loire region" C.P.E.R. "Innovations technologiques et Post-génomique". Sylvain Demey, a former student from the Master of Bioinformatics at Bordeaux, was assigned the task to improve and extend the platform, integrate all previous components in a software suite, homogenize the interfaces, implement other functionalities. This, he had to achieve between avril 2005 and december 2005. Not the least difficult of his tasks were among the following: debug, test and integrate code written by five other people; improve the client-server architecture, improve the session management, pipeline software components to provide an easy-to-use tool; implement the inescapable so-called "user-friendly" interface. By the end of year 2005, the platform had been tested with 45 genomes. Year 2006 was also favourable to BacTrans² project since it was granted 9 supplementary months by Ouest-Genopole consortium (from january 2006 to september 2006).
  
  As a conclusion, BacTrans² is now an enriched software suite. It yields the list of genes with putative strong promoters harbouring an UP element. It provides an "intra" statistics component to derive trends regarding strong transcription or strong translation over each genome selected. Besides, correlation coefficients are systematically computed and the criteria concerned are plotted one against each other. BacTrans² also provides various comparative genomics tools: an "inter" statistics component highlights the variations of the aptitude to strong expression through genomes; Position-Specific Scoring Matrices are computed for -10 and -35 boxes as well as UP element harboured in putative strong promoters; homologous genes (associated with strong promoters) are identified between any two genomes selected. Regarding the interface, it is still not known whether it is friendly or not with the user (at least, it is never rude); anyway, clickable links should favour an easy navigation through the software suite as well as enable some quick appropriate incursions to NCBI site.
  The platform is currently under limited access due to a lack of resources for maintenance.
  Some snapshots of the platform may be found  here (in French).