MBMichael Genot Benoit Blum

Michael Genot Benoit Blum

*This is an algorithmically generated author record based on publicly available information.

Michael Genot Benoit Blum has a total of 356 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Clinical Evaluation of an AI System for Streamlined Variant Interpretation in Genetic Testing

    by Jiri Ruzicka, Jean-Marie Ravel, Jérôme Audoux, Alexandre Boulat, Julien Thévenon, Kévin Yauy, Marine Dancer, Laure Raymond, Yannis Lombardi, Nicolas Philippe, et al.

    The growing use of genomic sequencing to diagnose hereditary diseases has increased the interpretive workload for clinical laboratories. Efficient methods are needed to maximize diagnostic yield without overwhelming resources. We developed DiagAI, an integrative machine-learning system trained to prioritize and sort causal variants in rare diseases. DiagAI integrates Universal Pathogenicity Predictor (UP 2 ), a machine-learning model trained to predict ACMG pathogenicity classes, PhenoGenius to match genotype-phenotype interactions and expert features such as inheritance and variant quality. We retrospectively analyzed 196 diagnosed exomes from a nephrology cohort. To benchmark UP 2 ’s performance, we evaluated the ranking of 62 causal missense variants. UP 2 ranked variants most effectively beyond shortlist sizes of 10 and identified pathogenic variants missed by AlphaMissense. DiagAI identified 94.9% of causal variants in diagnostic exomes with HPO terms, compared to 90.8% without, with median shortlist sizes of 12 and 9 variants, respectively. With HPO terms, 74% of top-ranked variants were diagnostic, versus 42% without, outperforming Exomiser and AI-MARRVEL. DiagAI produces accurate shortlists that streamline variant interpretation, offering a scalable solution for growing diagnostic volumes.


  • Open Access

    Robust Evaluation of Deep Learning-based Representation Methods for Survival and Gene Essentiality Prediction on Bulk RNA-seq Data

    by Baptiste Gross, Antonin Dauvin, Vincent Cabeli, Virgilio Kmetzsch, Jean El Khoury, Gaëtan Dissez, Khalil Ouardini, Simon Grouard, Alec Davi, Regis Loeb, et al.

    Deep learning (DL) has shown potential to provide powerful representations of bulk RNA-seq data in cancer research. However, there is no consensus regarding the impact of design choices of DL approaches on the performance of the learned representation, including the model architecture, the training methodology and the various hyperparameters. To address this problem, we evaluate the performance of various design choices of DL representation learning methods using TCGA and DepMap pan-cancer datasets, and assess their predictive power for survival and gene essentiality predictions. We demonstrate that non DL-based baseline methods achieve comparable or superior performance compared to more complex models on survival predictions tasks. DL representation methods, however, are the most efficient to predict the gene essentiality of cell lines. We show that auto-encoders (AE) are consistently improved by techniques such as masking and multi-head training. Our results suggest that the impact of DL representations and of pre-training are highly task- and architecture-dependent, highlighting the need for adopting rigorous evaluation guidelines. These guidelines for robust evaluation are implemented in a pipeline made available to the research community.


  • Open Access

    Clinical evaluation of a low-coverage whole-genome test for detecting homologous recombination deficiency in ovarian cancer

    by Romain Boidot, Michael G.B. Blum, Marie-Pierre Wissler, Céline Gottin, Jiri Ruzicka, Sandy Chevrier, Tiffany M. Delhomme, Jérome Audoux, Adrien Jeanniard, Pierre-Alexandre Just, et al.

    Background: The PAOLA-1/ENGOT-ov25 trial showed that maintenance olaparib plus bevacizumab increases survival of advanced ovarian cancer patients with homologous recombination deficiency (HRD). However, decentralized solutions to test for HRD in clinical routine are scarce. The goal of this study was to retrospectively validate on tumor samples from the PAOLA-1 trial, a decentralized HRD test based on low coverage shallow Whole Genome Sequencing (sWGS). Methods: The study comprised 368 patients from the PAOLA-1 trial. The sWGS test was compared to the Myriad MyChoice HRD test (Myriad Genetics), and results were analyzed with respect to Progression-Free Survival (PFS). Results: We found a 95% concordance between the HRD status of the two tests (95% Confidence Interval (CI) 92%-97%). The Positive Percentage Agreement (PPA) of the sWGS test was 95% (95% CI; 90%-97%) like its Negative Percentage Agreement (NPA) (95% CI; 89%-98%). Only 1% (95% CI; 0-3%) of its results were inconclusive. In patients with HRD-positive tumors treated with olaparib plus bevacizumab, the PFS Hazard Ratio (HR) was 0.38 (95% CI; 0.26-0.54) with sWGS and 0.32 (95% CI; 0.22-0.45) with the Myriad assay. In patients with HRD-negative tumors, HR was 0.98 (95% CI; 0.68-1.41) and 1.05 (95% CI; 0.70-1.57) with sWGS and Myriad tests. Among patients with BRCA-wildtype tumors, those with HRD-positive tumors, benefited from olaparib plus bevacizumab maintenance, with HR of 0.48 (95% CI: 0.29-0.79) and of 0.38 (95% CI: 0.23 to 0.63) with sWGS and Myriad test. Conclusion: The SeqOne sWGS assay offers a clinically validated approach to detect HRD.

  • Open Access

    More efficient and inclusive time-to-event trials with covariate adjustment: a simulation study

    by Raphaëlle Momal, Paul Trichelair, Michael G.B. Blum, Félix Balazard

    Adjustment for prognostic covariates increases the statistical power of randomized trials. The factors influencing increase of power are well-known for trials with continuous outcomes. Here, we study which factors influence power and sample size requirements in time-to-event trials. We consider both parametric simulations and simulations derived from the TCGA cohort of hepatocellular carcinoma (HCC) patients to assess how sample size requirements are reduced with covariate adjustment. Simulations demonstrate that the benefit of covariate adjustment increases with the prognostic performance of the adjustment covariate (C-index) and with the cumulative incidence of the event in the trial. For a covariate that has an intermediate prognostic performance (C-index=0.65), the reduction of sample size varies from 1.7% when cumulative incidence is of 10% to 26.5% when cumulative incidence is of 90%. Broadening eligibility criteria usually reduces statistical power while our simulations show that it can be maintained with adequate covariate adjustment. In a simulation of HCC trials, we find that the number of patients screened for eligibility can be divided by 2.7 when broadening eligibility criteria. Last, we find that the Cox-Snell is a good approximation of the reduction in sample size requirements provided by covariate adjustment. This metric can be used in the design of time-to-event trials to determine sample size. Overall, more systematic adjustment for prognostic covariates leads to more efficient and inclusive clinical trials especially when cumulative incidence is large as in metastatic and advanced cancers. Key messages Covariate adjustment is a statistical technique that leverages prognostic scores within the statistical analysis of a trial. We study its benefits for time-to-event trials. Power gain achieved with covariate adjustment is determined by the prognostic performance of the covariate and by the cumulative incidence of events at the end of the follow-up period. Trials in indications with large cumulative incidence such as metastatic cancers can benefit from covariate adjustment to improve their statistical power. Covariate adjustment maintains statistical power in trials when eligibility criteria are broadened.

  • Open Access

    Genetic population structure across Brittany and the downstream Loire basin provides new insights on the demographic history of Western Europe

    by Isabel Alves, Joanna Giemza, Michael Blum, Carolina Bernhardsson, Stéphanie Chatel, Matilde Karakachoff, Aude Saint Pierre, Anthony F. Herzig, Robert Olaso, Martial Monteil, et al.

    European genetic ancestry originates from three main ancestral populations - Western hunter-gatherers, early European farmers and Yamnaya Eurasian herders - whose edges geographically met in present-day France. Despite its central role to our understanding of how the ancestral populations interacted and gave rise to modern population structure, the population history of France has remained largely understudied. Here, we analysed the high-coverage whole-genome sequences and genome-wide genotype profiles of respectively 856 and 3,234 present-day individuals from the northern half of France, and merged them with publicly available present-day and ancient Europe-wide genotype datasets. We also explored, for the first time, the whole-genome sequences of six mediaeval individuals (300-1100 CE) from Western France to gain insights into the genetic impact of what is commonly known as the Migration Period in Europe. We found extensive fine-scale population structure across Brittany and the downstream Loire basin, emphasising the need for investigating local populations to better understand the distribution of rare and putatively deleterious variants across space. Overall, we observed an increased population differentiation between the northern and southern sides of the river Loire, which are characterised by different proportions of steppe vs. Neolithic-related ancestry. Samples from Western Brittany carry the largest levels of steppe ancestry and show high levels of allele sharing with individuals associated with the Bell Beaker complex, levels that are only comparable with those found in populations lying on the northwestern edges of Europe. Together, our results imply that present-day individuals from Western Brittany retain substantial legacy of the genetic changes that occurred in Northwestern Europe following the arrival of the Bell Beaker people c. 2500 BCE. Such genetic legacy may explain the sharing of disease-related alleles with other present-day populations from Western Britain and Ireland.


  • Open Access

    External control arm analysis: an evaluation of propensity score approaches, G-computation, and doubly debiased machine learning

    by Nicolas Loiseau, Paul Trichelair, Maxime He, Mathieu Andreux, Mikhail Zaslavskiy, Gilles Wainrib, Michael G.B. Blum

    Background: An external control arm is a cohort of control patients that are collected from data external to a single-arm trial. To provide an unbiased estimation of efficacy, the clinical profiles of patients from single and external arms should be aligned, typically using propensity score approaches. There are alternative approaches to infer efficacy based on comparisons between outcomes of single-arm patients and machine-learning predictions of control patient outcomes. These methods include G-computation and Doubly Debiased Machine Learning (DDML) and their evaluation for ECA analysis is insufficient. Methods: We consider both numerical simulations and a trial replication procedure to evaluate the different statistical approaches: propensity score matching, Inverse Probability of Treatment Weighting (IPTW), G-computation, and DDML. The replication study relies on five type 2 diabetes randomized clinical trials granted by the Yale University Open Data Access (YODA) project. From the pool of five trials, observational experiments are artificially built by replacing a control arm from one trial by an arm originating from another trial and containing similarly-treated patients. Results: Among the different statistical approaches, numerical simulations show that DDML has the smallest bias followed by G-computation. Ranking based on mean square error is different with G-computation always being among the lowest-error methods while DDML relative performance improves with increasing sample sizes. For hypothesis testing, DDML controls type-1 error and is conservative whereas G-computation and propensity score approaches can be liberal with type I errors ranging between 5% and 10% in some settings. G-computation is the best method in terms of statistical power, and DDML has comparable power at n = 1000 but its power is inferior to propensity score approaches at n = 250. The replication procedure also indicates that G-computation minimizes mean squared error while DDML has intermediate performances compared to G-computation and propensity score approaches. The confidence intervals of G-computation are the narrowest in lines with its liberal type I error whereas confidence intervals of DDML are the widest that confirms its conservative nature. Conclusions: For external control arm analyses, methods based on outcome prediction models can reduce estimation error and increase statistical power compared to propensity score approaches.

  • Top Cited Papers
    Open Access

    Integrating deep learning CT-scan model, biological and clinical variables to predict severity of COVID-19 patients

    by Nathalie Lassau, Samy Ammari, Emilie Chouzenoux, Hugo Gortais, Paul Herent, Matthieu Devilder, Samer Soliman, Olivier Meyrignac, Marie-Pauline Talabard, Jean-Philippe Lamarque, et al.

    The SARS-COV-2 pandemic has put pressure on intensive care units, so that identifying predictors of disease severity is a priority. We collect 58 clinical and biological variables, and chest CT scan data, from 1003 coronavirus-infected patients from two French hospitals. We train a deep learning model based on CT scans to predict severity. We then construct the multimodal AI-severity score that includes 5 clinical and biological variables (age, sex, oxygenation, urea, platelet) in addition to the deep learning model. We show that neural network analysis of CT-scans brings unique prognosis information, although it is correlated with other markers of severity (oxygenation, LDH, and CRP) explaining the measurable but limited 0.03 increase of AUC obtained when adding CT-scan information to clinical variables. Here, we show that when comparing AI-severity with 11 existing severity scores, we find significantly improved prognosis performance; AI-severity can therefore rapidly become a reference scoring approach.


  • Taste perception and lifestyle: insights from phenotype and genome data among Africans and Asians

    by Agnès E. Sjöstrand, Per Sjödin, Tatyana Hegay, Anna Nikolaeva, Farhad Shayimkulov, Michael G. B. Blum, Evelyne Heyer, Mattias Jakobsson

    Taste is essential for the interaction of animals with their food and has co-evolved with diet. Humans have peopled a large range of environments and present a wide range of diets, but little is known about the diversity and evolution of human taste perception. We measured taste recognition thresholds across populations differing in lifestyles (hunter gatherers and farmers from Central Africa, nomad herders, and farmers from Central Asia). We also generated genome-wide genotype data and performed association studies and selection scans in order to link the phenotypic variation in taste sensitivity with genetic variation. We found that hunter gatherers have lower overall sensitivity as well as lower sensitivity to quinine and fructose than their farming neighbors. In parallel, there is strong population divergence in genes associated with tongue morphogenesis and genes involved in the transduction pathway of taste signals in the African populations. We find signals of recent selection in bitter taste-receptor genes for all four populations. Enrichment analysis on association scans for the various tastes confirmed already documented associations and revealed novel GO terms that are good candidates for being involved in taste perception. Our framework permitted us to gain insight into the genetic basis of taste sensitivity variation across populations and lifestyles.


  • Open Access

    AI-based multi-modal integration (ScanCov scores) of clinical characteristics, lab tests and chest CTs improves COVID-19 outcome prediction of hospitalized patients

    by Nathalie Lassau, Samy Ammari, Emilie Chouzenoux, Hugo Gortais, Paul Herent, Matthieu Devilder, Samer Soliman, Olivier Meyrignac, Marie-Pauline Talabard, Jean-Philippe Lamarque, et al.

    The SARS-COV-2 pandemic has put pressure on Intensive Care Units, and made the identification of early predictors of disease severity a priority. We collected clinical, biological, chest CT scan data, and radiology reports from 1,003 coronavirus-infected patients from two French hospitals. Among 58 variables measured at admission, 11 clinical and 3 radiological variables were associated with severity. Next, using 506,341 chest CT images, we trained and evaluated deep learning models to segment the scans and reproduce radiologists’ annotations. We also built CT image-based deep learning models that predicted severity better than models based on the radiologists’ reports. Finally, we showed that adding CT scan information—either through radiologist lesion quantification or through deep learning—to clinical and biological data, improves prediction of severity. These findings show that CT scans contain novel and unique prognostic information, which we included in a 6-variable ScanCov severity score.


  • Open Access

    Efficient toolkit implementing best practices for principal component analysis of population genetic data

    by Florian Privé, Keurcien Luu, Michael G B Blum, John J McGrath, Bjarni J Vilhjálmsson

    Principal component analysis (PCA) of genetic data is routinely used to infer ancestry and control for population structure in various genetic analyses. However, conducting PCA analyses can be complicated and has several potential pitfalls. These pitfalls include (i) capturing linkage disequilibrium (LD) structure instead of population structure, (ii) projected PCs that suffer from shrinkage bias, (iii) detecting sample outliers and (iv) uneven population sizes. In this work, we explore these potential issues when using PCA, and present efficient solutions to these. Following applications to the UK Biobank and the 1000 Genomes project datasets, we make recommendations for best practices and provide efficient and user-friendly implementations of the proposed solutions in R packages bigsnpr and bigutilsr. For example, we find that PC19–PC40 in the UK Biobank capture complex LD structure rather than population structure. Using our automatic algorithm for removing long-range LD regions, we recover 16 PCs that capture population structure only. Therefore, we recommend using only 16–18 PCs from the UK Biobank to account for population structure confounding. We also show how to use PCA to restrict analyses to individuals of homogeneous ancestry. Finally, when projecting individual genotypes onto the PCA computed from the 1000 Genomes project data, we find a shrinkage bias that becomes large for PC5 and beyond. We then demonstrate how to obtain unbiased projections efficiently using bigsnpr. Overall, we believe this work would be of interest for anyone using PCA in their analyses of genetic data, as well as for other omics data. R packages bigsnpr and bigutilsr can be installed from either CRAN or GitHub (see https://github.com/privefl/bigsnpr). A tutorial on the steps to perform PCA on 1000G data is available at https://privefl.github.io/bigsnpr/articles/bedpca.html. All code used for this paper is available at https://github.com/privefl/paper4-bedpca/tree/master/code. Supplementary data are available at Bioinformatics online.