Reconstruction of large scale gene regulatory networks (GRNs in the following) is an important step for understanding the complex regulatory mechanisms within the cell. Many modeling approaches have been introduced to find the causal relationship between genes using expression data. However, they have been suffering from high dimensionality-large number of genes but a small number of samples, overfitting, heavy computation time and low interpretability. We have previously proposed an algorithm LICORN, which uses the discrete expression data to find cooperative regulation relationships that are out of the scope of most GRN inference methods. However, as many other methods, LICORN suffers from a large number of false positives. We propose here a hybrid inference method H-Licorn that combines Licorn with a numerical selection step, expressed as a linear regression problem, that effectively complements the discrete search of Licorn. We evaluate a bootstrapped version of H-LICORN on the in silico DREAM5 dataset and show that H-LICORN has significantly higher performance than LICORN, and is competitive or outperforms state of the art GRN inference algorithms, especially when operating on small data sets. We also applied H-LICORN on a real dataset of human bladder cancer and show that it performs better than other methods in finding candidate regulatory interactions. In particular, solely based on gene expression data, H-LICORN is able to identify experimentally validated regulator cooperative relationships involved in cancer.

Reconstruction of gene regulatory networks (GRNs) is an important step for understanding the complex regulatory mechanisms within the cell. Many modeling approaches have been introduced to find the causal relationship between genes using expression data. However, they have been suffering from high dimensionality - large number of genes but a small number of samples -, over fitting, and heavy computation time. In this work 1, we present a novel method, namely SETNET, to improve the stability and accuracy of GRN inference using ensemble techniques. For a given target gene, SETNET extract an ensemble of regulation networks from discretized expression data instead of a single one. Inferred networks are then assessed by ranking individual regulation relationships using a regression based technique and continuous expression data. Evaluation on DREAM5 data demonstrates that SETNET is efficient, specially when operating on a small data set.

SUMMARY: VIRTLAB is a self-training program based on PBL (Problem-Based-Learning Pathway) built to simulate a molecular biology laboratory. It has been designed to stimulate students in the biological sciences to analyse and solve molecular biology problems using standard laboratory techniques (e.g. restriction enzyme digestions, analytical and preparative agarose gels, DNA cloning and sequencing, etc.) and can thus be viewed as a teaching aid. AVAILABILITY: The VIRTLAB package is distributed free of charge to non-profit organisations by the authors ([email protected]. unina.it). On-line help and tutorials, available now in English, French, Italian, and shortly in German, are provided with the software or at http://biol.dgbm.unina.it:8080/virtlab.html++ +