Assessing NGS-based computational methods for predicting transcriptional regulators with query gene sets

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Accurate identification of transcriptional regulators (TRs) is crucial for understanding biological development, disease mechanisms, and therapeutic target prediction.
• Numerous computational methods utilizing next-generation sequencing (NGS) data exist, but a systematic evaluation is lacking.

Approach:

• Classified NGS-based TR prediction methods into library-based and region-based categories.
• Conducted benchmark studies evaluating accuracy, sensitivity, coverage, and usability of these methods using molecular experimental datasets.
• Compared NGS-based methods against traditional motif-based approaches.

Key Points:

• NGS-based methods generally outperform motif-based methods for TR prediction.
• Region-centric NGS methods utilizing larger databases exhibit better performance.
• BART, ChIP-Atlas, and Lisa are recommended for their strong performance across various scenarios.

Conclusions:

• Identified limitations of current NGS-based TR prediction tools.
• Highlighted potential areas for future improvements in computational TR identification methods.