A penalized likelihood approach for robust estimation of isoform expression

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Ultra-high-throughput sequencing of transcriptomes (RNA-Seq) allows gene expression estimation at the isoform level.
• However, systematic biases from sequencing, mapping, and incomplete annotation databases can lead to inaccurate isoform abundance estimates.

Purpose of the Study:

• To introduce a robust penalized likelihood approach for detecting and correcting biases in RNA-Seq data.
• To extend existing models by incorporating bias parameters for reads and utilizing an L1 penalty for parameter selection.

Main Methods:

• Development of a penalized likelihood model to address RNA-Seq isoform abundance estimation biases.
• Implementation of an efficient algorithm for model fitting and analysis of statistical properties.
• Application of an L1 penalty for the selection of non-zero bias parameters.

Main Results:

• The proposed model demonstrates potential in improving the accuracy of isoform-specific gene expression estimates.
• The method can effectively identify incompletely annotated gene models.
• Experimental studies on simulated and real datasets validate the model's performance.

Conclusions:

• The penalized likelihood approach offers a robust solution for bias correction in RNA-Seq.
• This method enhances the reliability of isoform abundance quantification.
• The approach aids in the discovery and refinement of gene annotations.