GRN Inference Algorithms

Arboreto hosts multiple (currently 2, contributions welcome!) algorithms for inference of gene regulatory networks from high-throughput gene expression data, for example single-cell RNA-seq data.

GRNBoost2

GRNBoost2 is the flagship algorithm for gene regulatory network inference, hosted in the Arboreto framework. It was conceived as a fast alternative for GENIE3, in order to alleviate the processing time required for larger datasets (tens of thousands of observations).

GRNBoost2 adopts the GRN inference strategy exemplified by GENIE3, where for each gene in the dataset, the most important feature are a selected from a trained regression model and emitted as candidate regulators for the target gene. All putative regulatory links are compiled into one dataset, representing the inferred regulatory network.

In GENIE3, Random Forest regression models are trained.

GENIE3

We consider GENIE3 as the blueprint of “multiple regression GRN inference” strategy.

DREAM5 benchmark