NetProphet 3: a machine learning framework for transcription factor network mapping and multi-omics integration.

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Mapping transcription factor (TF) targets from gene expression data is crucial for understanding gene regulation.
• Existing methods for combining TF target predictions often use simplistic or non-generalizable formulas.
• Accurate evaluation of TF target prediction methods is hindered by the lack of genome-scale, ground-truth networks.

Purpose of the Study:

• To develop an automated method for combining TF target predictions from multiple analyses.
• To establish robust, genome-scale evaluation metrics for TF target prediction.
• To create a consensus network integrating gene expression and TF binding location data.

Main Methods:

• Developed NetProphet3, a novel tool employing a tree boosting algorithm.
• Trained the algorithm on TF binding location data for enhanced prediction accuracy.
• Established three independent, genome-scale metrics for evaluating prediction accuracy.

Main Results:

• NetProphet3 demonstrates superior accuracy compared to existing packages, including NetProphet 2.0, when using direct TF perturbation gene expression data.
• The developed evaluation metrics provide a reliable assessment of TF target prediction performance.
• NetProphet3's integration mode successfully generates a consensus network by combining gene expression and TF binding data.

Conclusions:

• NetProphet3 offers a significant advancement in accurately mapping transcription factor targets.
• The tool provides a robust framework for integrating diverse biological data types for network inference.
• Availability of code and data facilitates reproducibility and further research in the field.