Practical aspects of gene regulatory inference via conditional inference forests from expression data.

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Gene regulatory network (GRN) inference is crucial for understanding complex biological interactions.
• Existing methods for GRN inference have limitations in performance and efficiency.

Purpose of the Study:

• To propose a novel framework for GRN inference using Conditional Inference Forests (CIFs).
• To evaluate the performance of CIF-based methods against existing approaches like Random Forests (RF).

Main Methods:

• Utilized ensembles of Conditional Inference Trees (CITs) within the CIF framework.
• Implemented aggregation schemes for variant selection prior to network construction.
• Compared CIFcond (conditional permutation) and CIFmean (P-value averaging) with RF on synthetic and DREAM4 challenge data.

Main Results:

• CIFcond demonstrated superior performance compared to RF on synthetic microarray data.
• CIFcond outperformed other CIF-based methods and scenarios in the DREAM4 challenge.
• CIFmean offered a balance of adequate performance and computational efficiency, with automatic thresholding.

Conclusions:

• CIF methodology provides a flexible and effective framework for inferring biological interactions.
• CIFmean presents a practical and efficient approach for GRN inference without parameter tuning.
• The study confirmed a biologically relevant interaction between IL2RA and FOXP1, relevant to IL-2 signaling and type 1 diabetes.