Evaluating Transcription Factor Activity Changes by Scoring Unexplained Target Genes in Expression Data.

Area of Science:

• Systems Biology
• Computational Biology
• Genomics

Background:

• Predicting transcription factor (TF) activity from gene expression data and regulatory networks is crucial but challenging.
• Existing gene regulatory networks are often incomplete and contain irrelevant condition-specific regulations.
• A systematic method to evaluate the accuracy of inferred TF activity changes is lacking.

Purpose of the Study:

• To develop and present a novel evaluation strategy for inferred TF activity changes.
• To quantify the number of target genes whose expression changes can be explained by active TFs.
• To introduce the inconsistency score (i-score) for evaluating unexplained gene expression changes.

Main Methods:

• Developed an evaluation strategy assuming a gene is explained if any TF combination can account for its observed expression change.
• Introduced the i-score to quantify unexplained target genes.
• Implemented optimization methods (Act-SAT, Act-A*) to calculate minimal i-scores and identify optimal TF activity sets.

Main Results:

• Published methods consistently result in a high number of unexplained target genes (large i-scores) across various datasets.
• Optimization methods reveal that even minimal i-scores indicate numerous unexplained target genes, suggesting incomplete regulatory networks.
• Act-SAT and Act-A* provide optimal TF activity sets for analyzing expression and network data.

Conclusions:

• Current gene regulatory networks are significantly incomplete, as evidenced by the high i-scores obtained even with optimized methods.
• The developed i-score and associated optimization tools provide a robust framework for evaluating TF activity inference.
• The findings underscore the need for more comprehensive and accurate gene regulatory network reconstructions.