Repeat or not repeat?--Statistical validation of tandem repeat prediction in genomic sequences.

Area of Science:

• Genomics
• Bioinformatics

Background:

• Tandem repeats (TRs) are abundant and functionally significant genomic features.
• Despite numerous tools, accurate TR detection remains an unresolved challenge.
• Current algorithms produce diverse and often conflicting TR predictions.

Purpose of the Study:

• To evaluate the performance of existing TR detection tools.
• To develop a statistical framework for assessing the quality of predicted TRs.
• To improve the accuracy and completeness of TR identification in genomic data.

Main Methods:

• Large-scale testing of current TR detection algorithms.
• Simulations to analyze the impact of repeat characteristics on detection power.
• Development and evaluation of a model-based phylogenetic classifier for TR divergence.
• Application of statistical criteria to filter false-positive TR predictions.

Main Results:

• TR detection accuracy is influenced by repeat divergence and unit characteristics.
• Current detectors exhibit algorithm-specific biases, leading to non-overlapping inferences.
• The proposed statistical classification scheme effectively filters false positives.
• Combining multiple detectors with filtering enhances the identification of genuine TRs.

Conclusions:

• TR detection requires careful consideration of algorithm-specific strengths and weaknesses.
• A statistical classification approach is crucial for reconciling diverse predictions.
• Applying multiple TR detectors followed by filtering is recommended for comprehensive and accurate results.