Publications
How to organise a scientific competition to benchmark methods and algorithms in computational biology?
HAL
This paper provides a comprehensive guide for organizing scientific competitions in bioinformatics, based on our experience with HADACA3, a data challenge focused on deconvolution algorithms for predicting cellular composition in cancer, from multi-omics data.
A robust workflow to benchmark deconvolution of multi-omic data
Genome Biology
This manuscript presents a comprehensive and unbiased evaluation framework for benchmarking deconvolution algorithms across transcriptomic and methylomic data, addressing critical gaps in existing studies and providing key advances in the quantification of tumor heterogeneity from bulk molecular data.
DNA methylation and immune infiltration mediate the impact of tobacco exposure on pancreatic adenocarcinoma outcome: a high-dimensional mediation analysis
BioRxiv
In this work, we developed HDMAX2-surv, a novel framework for high-dimensional mediation analysis specifically adapted to censored survival data. Our approach integrates computational immune deconvolution with causal discovery and serial mediation analysis, addressing a critical methodological gap in understanding how molecular intermediates shape clinical outcomes.
Group lasso based selection for high-dimensional mediation analysis
Statistics in Medicine
This paper presents a two-step procedure for high-dimensional mediation analysis. The first step selects a reduced number of candidate mediators using an ad-hoc lasso penalty. The second step applies a procedure we previously developed to estimate the mediated and direct effects, accounting for the correlation structure among the retained candidate mediators.
Redefining phenotypic intratumor heterogeneity of pancreatic ductal adenocarcinoma: a bottom-up approach
The Journal of Pathology
Herein, we developed a panel of antibodies that could easily be used by researchers and pathologists. The purpose of this panel was to classify patients according to the two main subtypes of PDAC, roughly basal-like or classical. To achieve this, we selected markers through a stringent and multistep process.
hdmax2, an R package to perform high dimension mediation analysis
Peer Community Journal
This manuscript introduces HDMAX2, a statistical method and R package developed to conduct mediation analysis in high-dimensional settings.
AI Competitions and Benchmarks, Practical issues: Proposals, grant money, sponsors, prizes, dissemination, publicity
DMLR
This book explains how AI competitions and benchmarks are created, run, and used. It brings together lessons from experienced organizers in academia, industry, and non-profits. Covering topics like datasets, evaluation, platforms, and incentives, it shows how challenges drive research, education, and innovation. Designed for researchers, engineers, and organizers, it is a practical guide to understanding and building impactful AI competitions : book URL
DECOMICS, a shiny application for unsupervised cell type deconvolution and biological interpretation of bulk omic data
Bioinformatics Advances
Our article presents a user-friendly Shiny application for estimating and identifying cell type composition from bulk transcriptomes using unsupervised approaches. Additionally, the application offers guidance for conducting analyses and interpreting the biological implications of the results.
Pacpaint: a histology-based deep learning model uncovers the extensive intratumor molecular heterogeneity of pancreatic adenocarcinoma
Nature Communications
To allow rapid PDAC molecular subtyping and study PDAC heterogeneity, we develop PACpAInt, a multi-step deep learning model. PACpAInt correctly predicts tumor subtypes at the whole slide level on surgical and biopsies specimens and independently predicts survival. PACpAInt highlights the presence of a minor aggressive Basal contingent that negatively impacts survival in 39% of RNA-defined classical cases.
Codabench: Flexible, easy-to-use, and reproducible meta-benchmark platform
Patterns
We introduce Codabench, a meta-benchmark platform, that is capable of flexible and easy benchmarking and supports reproducibility. Codabench is an important step toward benchmarking and reproducible research. It has been used in various communities including graph machine learning, cancer heterogeneity, clinical diagnosis, and reinforcement learning. Codabench is ready to help trendy research, e.g., artificial intelligence (AI) for science and data-centric AI.
DECONbench: a benchmarking platform dedicated to deconvolution methods for tumor heterogeneity quantification
BMC bioinformatics
Here we propose an innovative public digital benchmarking platform, open source, and freely available for the scientific community, including both high quality benchmarking datasets and reference computational methods. The platform can be used to assess the performance of newly developed methods, which are automatically compared to the existing ones in a user-friendly fashion.
Guidelines for cell-type heterogeneity quantification based on a comparative analysis of reference-free DNA methylation deconvolution software
BMC Bioinformatics
This manuscript compares three software packages that infer cell type proportions based on methylation data. We here evaluate key factors affecting performance of deconvolution pipelines. We examine to what extent cell-type proportions can be accurately inferred when accounting for measured confounding factors.
PenDA, a rank-based method for personalized differential analysis: Application to lung cancer
PLOS Computational Biology
This manuscript describes a novel method, named PenDA, to perform differential analysis of gene expression at the individual level. In PenDA, a gene is considered as deregulated in one sample of interest (e.g., a tumor) if its local ordering relatively to other genes with similar expressions is perturbed compared to its ordering in a set of control samples (e.g., normal tissues).
Assigning function to natural allelic variation via dynamic modeling of gene network induction
Molecular Systems Biology
CRELD1 is an evolutionarily-conserved maturational enhancer of ionotropic acetylcholine receptors
eLife
Genomics of cellular proliferation in periodic environmental fluctuations
Molecular Systems Biology
Exploiting Single-Cell Quantitative Data to Map Genetic Variants Having Probabilistic Effects
PLOS Genetics
Biosynthesis of ionotropic acetylcholine receptors requires the evolutionarily conserved ER membrane complex
Proceedings of the National Academy of Sciences