Transcription start site signal profiling improves transposable element RNA expression analysis at locus-level

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Transposable Elements (TEs) transcriptional activity is linked to neurodegenerative diseases like ALS and FTLD.
• Analyzing TE expression via short-read sequencing is difficult due to repetitive sequences and TE exaptation in RNA.
• Existing TE quantification tools use probabilistic redistribution or discard multimappers, with limited benchmarking at the individual loci level.

Purpose of the Study:

• To compare the performance of TE quantification tools (SQuIRE, TElocal, SalmonTE) and simplistic strategies (featureCounts) at the individual loci level.
• To evaluate false discovery rates and identify drivers of false positives in TE expression analysis.
• To develop improved methods for accurate TE expression profiling.

Main Methods:

• Comparison of TE quantification tools (SQuIRE, TElocal, SalmonTE) and featureCounts (unique, fraction, random modes) using simulated RNA-seq datasets.
• Analysis of false discovery rate (FDR) and false positive drivers.
• Application of filtering strategies (read counts, baseMean) and Transcription Start Site (TSS) mapping statistics with k-means clustering.

Main Results:

• All tested quantification strategies, including the best performer TElocal, exhibited high false discovery rates exceeding true positives.
• Filtering by minimum read counts or baseMean expression significantly improved F1 scores and reduced false positives.
• Profiling TSS mapping statistics with k-means clustering substantially enhanced TElocal's performance.

Conclusions:

• Current TE quantification methods, even advanced ones, suffer from high false discovery rates when analyzing individual loci.
• Filtering and TSS profiling are crucial for improving the accuracy and reliability of TE expression analysis.
• The developed approach offers a more robust method for detecting and quantifying differentially expressed TEs in RNA-seq data.