NIMEFI

NIMEFI infers gene regulatory networks from high-throughput gene expression data using multiple ensemble feature importance algorithms.

Key Features:

• Regression Decomposition Strategy: Implements a regression decomposition strategy similar to GENIE3 that decomposes GRN inference into individual regression problems predicting each target gene from all other genes.
• Ensemble Feature Importance Methods: Applies multiple ensemble feature importance algorithms across regression methods including support vector regression, elastic net, random forest regression, symbolic regression, and their ensemble variants.
• Subsampling Approach: Uses a subsampling technique to convert feature selection algorithms into ensemble feature importance methods by averaging feature importances across multiple subsampled models.
• Rankwise Averaged Predictions: Integrates rankwise averaged predictions from multiple ensemble algorithms to combine and aggregate results across methods.

Scientific Applications:

• Gene regulatory network inference: Inferring gene regulatory networks (GRNs) from high-throughput gene expression measurements.
• Prioritization of regulatory links: Prioritizing putative regulatory interactions by aggregating feature importance scores across ensemble models.
• Benchmark evaluation: Assessing and comparing GRN inference performance in community challenges such as DREAM4 and DREAM5.

Methodology:

Per-gene regression decomposition as in GENIE3; ensemble feature importance computed using support vector regression, elastic net, random forest regression, symbolic regression and their ensemble variants; subsampling to average feature importances across multiple models; and rankwise averaging of predictions across algorithms.