More Accurate Transcript Assembly via Parameter Advising.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Genomic analysis software is increasingly complex with numerous tunable parameters.
• Parameter choice significantly impacts the accuracy of genomic analysis results, particularly transcript assembly.
• Existing tools often rely on default parameters, which may not be optimal for specific datasets.

Purpose of the Study:

• To quantify the impact of parameter choice on transcript assembly accuracy.
• To develop a method for automatically selecting input-specific parameter values for reference-based transcript assembly.
• To improve the automation of genomic analysis pipelines.

Main Methods:

• Developed a method for automatic parameter value selection for the Scallop transcript assembly tool.
• Evaluated the method on 1595 RNA-Seq samples from the Sequence Read Archive.
• Assessed the generalizability of the approach by applying it to the StringTie tool using 65 ENCODE RNA-Seq experiments.

Main Results:

• Automatic parameter selection for Scallop increased the area under the receiver operator characteristic curve (AUC) by an average of 28.9% compared to default parameters.
• Applying the approach to StringTie improved AUC by an average of 13.1%.
• The developed method demonstrates a significant improvement in transcript assembly accuracy through automated parameter optimization.

Conclusions:

• Automated, input-specific parameter selection is effective in enhancing the accuracy of genomic analysis tools.
• This approach represents a step towards fully automated genomic analysis pipelines.
• Parameter advisors for Scallop and StringTie are publicly available to facilitate adoption.