I am analyzing paired-end RNA-seq data with the primary goal of studying alternative splicing. As part of the preprocessing pipeline, I performed quality control and adapter trimming using Trimmomatic.

Despite this, I still observe residual polyA (and occasionally polyT) stretches in a subset of reads. To address this, I am considering adding polyA/polyT sequences to the adapter file in Trimmomatic in order to remove these homopolymer tails.

However, I have concerns about the potential impact of this approach on downstream analyses, particularly alternative splicing detection.

Specifically, I would appreciate insights on the following points:

1. Given that Trimmomatic is primarily designed for adapter removal, is it appropriate to use it for polyA/polyT trimming by introducing these sequences as adapters?
2. Could this strategy lead to over-trimming, especially in A-rich regions (e.g., 3' UTRs), and thereby remove biologically relevant sequence information?
3. What are the potential consequences of such trimming on splice junction detection and isoform quantification?
4. In practice, is it preferable to leave short residual polyA tails untrimmed rather than risk introducing bias?

Any recommendations, experiences, or references would be highly appreciated.