BRANE power (Biologically-Related Apriori Network Estimation)

Publications with BRANE methodology (Biologically-Related Apriori Network Estimation):
  • BRANE HK (aka BRANE Cone): upcoming
Globally, we call that BRANE power. In a word, a series of optimization methodologies combined with biologically-related a-priori for genomic/transcriptomic data analysis and gene network inference: BRANE Cut, BRANE Clust, BRANE Relax, and more to come. They are developed at IFP Energies nouvelles, see The development of "omic" technologies: gene expression, digital style, or bioinformatics serving more sustainable chemistry.
BRANE: Biologically-related A priori Network Enhancement for gene regulation network
We are developing a suite of bioinformatics tools based on graphs and optimization, dedicated to -omics gene expression data (from RNA-seq or microarrays). They are meant for Gene Regulatory Network (GRN) inference and genomic data analysis. Genomic and transcription data is complicated. Especially, gene-condition and transcription factor-target gene ratios are not favorable. Data counts are even biased, noisy and variable.

There exists therefore a large number of solutions. One needs to introduce a priori (co-regulation,  co-expression, clustering, modularity, sparsity) to improve the graph structure, for instance around modules, or using gene clustering. We use biologically-plausible assumptions, and propose a couple of Biologically-Related A priori Network Enhancement (BRANE) techniques. They can be used in post-processing of other network inferences techniques like CLR (Context Likelihood of Relatedness) or GENIE3 (GEne Network Inference with Ensemble of trees). They have been successfully tested on DREAM4 and DREAM5 challenges, Escherichia coli, Trichoderma reesei. Graph inference in the inferred Escherichia coli network is evaluated using the STRING database. Clustering is compared to SIMONE, WGCNA, X-means and RegulonDB.

The BRANE approach notably uses image processing and computer vision algorithms (dual thresholding, proximal methods, graph cuts, maximization-majorization, variable metrics), as noted in Enhancing gene regulatory network inference through data integration with markov random fields, Banf & Rhee, Nature Scientific Reports, 2017.
Borrowing concepts from the field of Computer Vision to infer gene regulatory networks in prokaryotes has recently gained some attention
The related results are detailed in Aurélie Pirayre PhD Thesis (Defense: Monday 3 July 2017): Reconstruction and clustering with graph optimization and priors on gene networks and images (PhD manuscript and slides). Meanwhile, the following papers and codes are available.

BRANE application papers
A note in passing: on Gilles Azzoni, wine producer (Le Raisin et l'Ange), who inspired our BRANE tools with his biodynamic Bran wines.