Authors & Credits

Core Team

  • Edi Prifti – Creator and maintainer of the Predomics approach and package since 2015. He proposed the genetic algorithm approaches (TerGa1 and TerGa2), the concept of the Family of Best Models, numerous visualization tools, and the web application architecture. Corresponding author. ORCID: 0000-0001-8861-1305 · IRD / UMMISCO

  • Jean-Daniel Zucker – Co-creator who brought numerous founding ideas behind the method since 2015. He also proposed the beam search heuristic (TerBeam) and has co-led the work. ORCID: 0000-0002-5597-7922 · IRD / UMMISCO

  • Yann Chevaleyre – Involved in the initial developments and worked on the concepts of balances and the mathematical optimization method (TerDa). LIPN

  • Blaise Hanczar – Also involved in the initial developments and contributed notably to the selection of the best models and other general ideas behind the framework. IBISC

  • Eugeni Belda – Joined the project in 2019 and contributed to the interpretability of the signatures, notably applied to the human microbiome, as well as their visualization. He has also extensively tested the package using a plethora of datasets. ORCID: 0000-0003-4307-5072 · IRD

gpredomics Rust Engine Team

  • Louison Lesage – Lead Rust developer for the gpredomics engine rewrite, delivering up to 1,000x performance improvements over the original R implementation. ORCID: 0009-0000-0252-6311 · GMT Science

  • Raynald de Lahondès – Rust developer contributing to the gpredomics engine architecture and optimization. ORCID: 0009-0000-2862-9589 · GMT Science

  • Vadim Puller – Scientific developer contributing to algorithm implementation and validation in the Rust engine. ORCID: 0000-0002-3900-8283 · GMT Science

Contributors

  • Lucas Robin – Joined the project in 2016 as part of a student project and worked on the implementation of the TerGa2 algorithm and brought code optimization elements.
  • Shasha Cui – Joined the project as part of a student internship in 2017. Her work focused on the concepts of feature importance and model stability analysis.
  • Magali Cousin Thorez – Joined the project in 2019 during a student internship. Her work focused on the simplification of classification signatures and their exploration in the context of microbial ecosystems (Interpred).
  • Youcef Sklab – Co-led with Edi a couple of student projects in collaboration with the Sup Galilée engineering school to build the R Shiny PredomicsApp application.
  • Gaspar Roy – Worked in 2023 on an evolved version of the R Shiny application.
  • Fabien Kambu – Contributed to testing, documentation, and deployment of the PredomicsApp web application.

Funding

  • ANR DeepIntegrOmics (ANR-21-CE45-0030) – End-to-End Deep Learning for Precision Medicine through Metagenomics and Cost-Sensitive Data Integration. PI: Jean-Daniel Zucker (UMMISCO). Partners: UMMISCO, NUTRIOMICS, LAMSADE, IBISC.

Industry Partner

  • GMT Science (Paris, France) – Contributed the Rust development team for gpredomics and provides expertise in high-performance scientific computing.

Citation

If you use Predomics in your research, please cite:

Prifti E., Chevaleyre Y., Hanczar B., Belda E., Danchin A., Clément K., & Zucker J.-D. (2020). Interpretable and accurate prediction models for metagenomics data. GigaScience, 9(3), giaa010. doi:10.1093/gigascience/giaa010

This site uses Just the Docs, a documentation theme for Jekyll.