All our applications are free, open-source and require no registration. Each runs in the browser as a Shiny app, and each can also be installed and run locally in R.
Every app has a permanent L2IB address of the form https://www.l2ib-lab.fr/app/<app-name>/.
Cite that address in papers, never the hosting provider’s URL — it lets us change host without breaking published links.
Clinical prediction
AIPAL
Predicts the subtype of acute leukaemia — APL, ALL or AML — from ten routine laboratory parameters.
Distinguishing acute promyelocytic leukaemia from other subtypes is a therapeutic emergency, and the tests that settle it are not available everywhere within the hour. AIPAL uses only parameters produced by any routine laboratory, so it can triage immediately and flag the cases where confidence is high enough to act on.
▶ Launch the app Source code Publication
Research tool. AIPAL is intended for research and educational purposes. It does not constitute a medical device and must not replace clinical judgement or confirmatory diagnostic testing. No data entered in the application is stored by the L2IB team.
Alcazer V, et al. Lancet Digital Health 2024;6(5):e323–e333. doi:10.1016/S2589-7500(24)00044-X · PubMed
FIBOM-AI
Predicts grade 2–3 bone marrow fibrosis from a routine complete blood count.
Assessing marrow fibrosis normally requires a bone marrow biopsy. FIBOM-AI estimates the probability of significant fibrosis from data already available for every patient, and reports both a general prediction and a stricter “confident” prediction based on clinical cutoffs.
Research tool. FIBOM-AI is intended for research and educational purposes. It does not constitute a medical device and must not be used as a substitute for clinical judgement, a bone marrow biopsy, or the advice of a qualified practitioner. No data entered in the application is stored by the L2IB team.
Research software
PIO — Panel Informativity Optimizer
Designs and benchmarks cancer NGS panels: the smallest set of genes or exons that still covers the most patients.
Choosing the genes for a sequencing panel trades informativity — the share of patients carrying at least one mutation in the panel — against panel length, which drives both cost and sensitivity. PIO optimises that trade-off from patient-level mutational data, and can also assess or complete a panel you already use. The published evaluation reports panel size reductions of up to 1,000 kb at equal informativity.
▶ Launch the app Source code Publication
Alcazer V, Sujobert P. The Journal of Molecular Diagnostics 2022;24(6):697–709. doi:10.1016/j.jmoldx.2022.03.005 · PubMed
StatAid
A graphical interface for statistical analysis, for clinicians and researchers who would rather not write code.
StatAid walks through the steps of a sound analysis rather than exposing a bare toolbox, which makes it usable for teaching as well as for day-to-day clinical research. A quick-start guide is available in English and French in the repository.
▶ Launch the app Source code Publication
Alcazer V. Journal of Open Source Software 2020;5(55):2630. doi:10.21105/joss.02630
Running them locally
All four are R packages. In R:
install.packages("remotes")
remotes::install_github("VincentAlcazer/AIPAL"); AIPAL::run_app()
remotes::install_github("VincentAlcazer/FIBOM-AI"); FIBOMAI::run_app()
remotes::install_github("VincentAlcazer/StatAid"); StatAid::run_app()
remotes::install_github("VincentAlcazer/PIO") # see the repository vignetteAll are released under the MIT licence.
Adding a new app
Gabarit — déployez l’app, créez app/<nom>/index.qmd en copiant celui de FIBOM-AI et en changeant l’URL cible, puis copiez un bloc .app-card ci-dessus. L’app est alors citable de façon permanente à https://www.l2ib-lab.fr/app/<nom>/.