Systems Biology Tools
In this section you will find the link and general concepts of different tools that we are developing to help advance (Plant) Systems Biology.
2024:
We have been cooking (for ~8 years now) what we think could be a cool tool to access unseen signal in genetic data.
Here's some quick background. When a GWAS study is carried out, genetic markers (often SNPs) associated with traits or diseases are identified. However, even after these markers have been identified, it is often found that these markers together explain only a small fraction of the total heritability (variance) estimated for these traits. This is the notion of the "missing heritability".
Epistasis refers to interactions between different genetic variants, where, for example, the effect of one gene may depend on the presence or absence of other genes. These complex interactions are often not captured by standard GWAS analyses, which look for simple additive effects of individual genes. If traits are influenced by complex interactions between many genes (epistasis), "conventional" GWAS may miss these interactions because they are designed to detect independent, unitary effects of variants. The methodological barrier to this measurement is the fact that current GWAS statistical models would require years of computation to evaluate the entire epistatic combinatorial space for a single phenotype.
We then developed the Next-Gen GWAS (NGG) (Paper in Genome Biology --> here) (which relies on a mathematical trick combined with the high-performance computing of GPUs [Graphic Processing Units]) that now evaluates over >80 billion first-order combinatorial interactions in less than 2 hours (see Figure purple and yellow below). We have applied NGG to Arabidopsis thaliana, providing for the first time 2D epistatic maps at gene resolution. We demonstrate for several phenotypes that a large proportion (regularly doubled) of the missing heritability can be recovered, that it indeed resides in epistatic interactions, and that it can be used to improve phenotype prediction (check the whole story here). This technology is now available through BionomeeX and developped for other organisms.
TransDetect has been released in Plant Physiology [link]. From transcriptomic data TransDetect tries to identify a set of TFs that may interact in the control of a given gene.
Our algorithm named FRANK has been released in NPJ systems biology.
FRANK is a large Gene Regulatory Network (GRNs) simulator that helped us to define some potential emerging properties of GRNs, and to understand rules to learn them using transcriptomics and prior knowledge on GNR (here).
FRANK output. Simulated gene expression and Large GRN:
A "click and go"online version of the FRANK algorithm (GRN simulator) written by Clement Carre is available here --> [M2SB.org]
Our algorithm named GeneCloud has been released in Molecular Plant.
GeneCloud explores semantic enrichments in your gene lists. It is provided through a web platform enabling click and go analysis. This platform generates data compatible with Virtual Plant formats [Katari et al. 2009].
Go play with it! --> [M2SB.org]
GeneCloud output:
You can also download the code [here] developed by Dr Piotr Mirowski in collaboration with Pr Shasha and Pr Lecun, that modeled GRNs responding to Nitrate. This correspond to State Space Modeling techniques described in our Genome Biol paper [here].
I developed a program named GeneSect, in collaboration with Pr Dennis Shasha, that determines whether the overlap between two gene lists is higher/lower than expected by chance.
The procedure is described in this paper [here] and is very useful to study signal convergence. GeneSect is available through the Virtual Plant Platform hosted at NYU (Pr Gloria Coruzzi lab) [Katari et al. Plant Phys 2009, the link to Virtual Plant is here, chose Analysis/Genesect].
GeneSect has been extensivelly used in order to probe NLP7 and CHL1/NRT1.1 effect on Primary Nitrate Response in a meta-analysis published here.
GeneSect output:
Coming Soon
Regine is in development ;) [link]