BRANE power: Gene network inference with graph optimization

And the IPFEN 2018 Yves Chauvin PhD prize is awarded to Aurélie Pirayre, for IFPEN first thesis on bioinformatics with graph optimization for gene networks. In French, you can now read ‘‘BRANE Power’’ : gènes et algorithmes, une alliance pour la chimie verte.

Aurélie Pirayre, Grégoire Allaire, Didier Houssin, Pierre-Henri Bigeard, Eric Heintzé

PhD manuscript and slides for her thesis:

• Reconstruction and clustering with graph optimization and priors on gene networks and images (manuscript)
• Abstract: The discovery of novel gene regulatory processes improves the understanding of cell phenotypic responses to external stimuli for many biological applications, such as medicine, environment or biotechnologies. To this purpose, transcriptomic data are generated and analyzed from DNA microarrays or more recently RNAseq experiments. They consist in genetic expression level sequences obtained for all genes of a studied organism placed in different living conditions. From these data, gene regulation mechanisms can be recovered by revealing topological links encoded in graphs. In regulatory graphs, nodes correspond to genes. A link between two nodes is identified if a regulation relationship exists between the two corresponding genes. Such networks are called Gene Regulatory Networks (GRNs). Their construction as well as their analysis remain challenging despite the large number of available inference methods. In this thesis, we propose to address this network inference problem with recently developed techniques pertaining to graph optimization. Given all the pairwise gene regulation information available, we propose to determine the presence of edges in the final GRN by adopting an energy optimization formulation integrating additional constraints. Either biological (information about gene interactions) or structural (information about node connectivity) a priori have been considered to restrict the space of possible solutions. Different priors lead to different properties of the global cost function, for which various optimization strategies, either discrete and continuous, can be applied. The post-processing network refinements we designed led to computational approaches named BRANE for \Biologically-Related A priori for Network Enhancement". For each of the proposed methods --- BRANE Cut, BRANE Relax and BRANE Clust --- our contributions are threefold: a priori-based formulation, design of the optimization strategy and validation (numerical and/or biological) on benchmark datasets from DREAM4 and DREAM5 challenges showing numerical improvement reaching 20%. In a ramification of this thesis, we slide from graph inference to more generic data processing such as inverse problems. We notably invest in HOGMep, a Bayesian-based approach using a Variational Bayesian Approximation framework for its resolution. This approach allows to jointly perform reconstruction and clustering/segmentation tasks on multi-component data (for instance signals or images). Its performance in a color image deconvolution context demonstrates both quality of reconstruction and segmentation. A preliminary study in a medical data classification context linking genotype and phenotype yields promising results for forthcoming bioinformatics adaptations.
• Slides (PhD defense on July 3rd, 2017)

are finally online (check the EURASIP Library of Ph.D. Theses). The work was ruled by the concept of BRANE power; a methodology for gene regulatory network inference and clustering based on graph optimization and biological priors. BRANE stands for Biologically Related Apriori Network Enhancement. It rhymes with cell membrane (and brain, for who it's worth).

Gene regulatory network inference with BRANE Cut

State-of-the-art results are obtained on synthetic and real transcriptomic data (DREAM-4, DREAM-5 for DREAM consortium challenges, Escherichia coli dataset). Derived methods are BRANE Cut (with graph cuts), BRANE Relax (with proximal optimization) and BRANE Clust (with graph Laplacian). 

Gene network joint inference and clustering with BRANE Clust

Another rule is termed HOGMep, with HOGMep: variational Bayes and higher-order graphical models applied to Joint Image Recovery and Segmentation.

Used concepts include:

• data science, optimization on graphs: maximal flow, minimum cut, random walker algorithm, variational and Bayes variational formalism, convex relaxation, alternating optimization, combinatorial Dirichlet problem, hard-clustering and soft-clustering
• biology, biotechnology, bioinformatics: transcription factors (TFs) as regulators and non-transcription factors (TFs) as targets, modular networks, biological priors, in-silico data, second generation bio-fuel production, DREAM4 challenge, DREAM5 challenge
• use to biofuels and green chemistry production (with fungus Trichoderma reesei)

Supervising team:

• Frédérique BIDARD-MICHELOT, IFP Energies nouvelles
• Camille COUPRIE, Facebook A.I. Research
• Laurent DUVAL, IFP Energies nouvelles
• Jean-Christophe PESQUET, CentraleSupélec

PhD Thesis reporters

• Pascal FROSSARD, EPFL (Ecole Polytechnique Fédérale de Lausanne), Switzerland
• Jean-Philippe VERT, Mines ParisTech, Paris, France

PhD Thesis Examiners

• Stéphane ROBIN, MIA Paris, INRA
• Hugues TALBOT, Université Paris-Est, Institut Gaspard Monge

More links:

• Research gate: Reconstruction and clustering with graph optimization and priors on gene networks and images
• BRANE power: a page for BRANE power avatars
• ResearchGate
• These.fr: Reconstruction et classification par optimisation dans des graphes avec a priori pour les réseaux de gènes et les images

By the way, brain (and evolution) is celebrated in the 44th France Culture jingles compilation, Kultur Pop.

A Primal-Dual Proximal Algorithm for Sparse Template-Based Adaptive Filtering: Application to Seismic Multiple Removal

A year ago we talked about a technique for Adaptive multiple subtraction with wavelet-based complex unary Wiener filters. The field of application is seismic signal processing. The fast and simple design was heuristic (helping discovery, stimulating interest as a means of furthering investigation.based on experimentation), based on an appropriate combination of "a sparsifying transform" and a closed-form one-tap, sliding-window adaptive filter. To make it more pragmatic (based on observation and real-world models), an alternative approach uses proximal algorithms to incorporate sparsity priors, either on data in redundant frame transforms and in the short-support filter design. Here is the preprint: A Primal-Dual Proximal Algorithm for Sparse Template-Based Adaptive Filtering: Application to Seismic Multiple Removal and the version published by IEEE Transactions on Signal Processing.

Unveiling meaningful geophysical information from seismic data requires to deal with both random and structured ``noises''. As their amplitude may be greater than signals of interest (primaries), additional prior information is especially important in performing efficient signal separation. We address here the problem of multiple reflections, caused by  wave-field bouncing between layers. Since only approximate models of these phenomena are available, we propose a flexible framework for time-varying adaptive filtering of seismic signals, using sparse representations,  based on inaccurate templates. We recast the joint estimation of adaptive filters and primaries in a new convex variational formulation. This approach allows us to incorporate plausible knowledge about noise statistics, data sparsity and slow filter variation in parsimony-promoting wavelet frames.  The designed primal-dual algorithm solves a  constrained  minimization problem that alleviates standard regularization issues in finding hyperparameters. The approach demonstrates  significantly good performance in low signal-to-noise ratio conditions, both for simulated and real field data.

While applied here to seismic signal, the concept is heavily related to pattern matching in images, echo cancellation in audio or voice signals, exemplar search in speech.