Pre-Trimming With Cutadapt when using the STAR Mapper

Recommended pre-trimming for DRAGEN v4.5 (DRAGEN-STAR workflow)

The standard DRAGEN hardware mode (run with FPGA acceleration) trims reads automatically. However, when processing scRNA-seq data with DRAGEN v4.5 software mode, using the DRAGEN-STAR workflow, trimming is not performed. We recommend pre-trimming FASTQs to remove TSO, polyA, and polyG sequences that can negatively impact alignment (leading to a variable drop in mapping rates, depending on the sample type and read 2 length). Note: Trimming should only be performed on gene expression libraries. TSO sequences are typically observed in <1–5% of reads from Illumina Single-Cell Prep Libraries and are more common in shorter, unfragmented library molecules. PolyA sequences occur when reads extend into the polyA tail or reverse-complement polyT capture sequence. PolyG sequences arise from reads containing short inserts, which generate dark cycles after processing past the end of the library molecule. Dark cycles are as G bases on Illumina instruments. Manual pre-trimming is recommended until support for these artifacts is incorporated in a future DRAGEN release (v4.6). The workflow below produces trimmed FASTQ pairs while trimming only Read 2, allowing the output files to be used directly with the DRAGEN-STAR pipeline.The standard DRAGEN hardware mode (run with FPGA acceleration) trims reads automatically. However, when processing scRNA-seq data with DRAGEN v4.5 software mode, using the DRAGEN-STAR workflow, trimming is not performed. We recommend pre-trimming FASTQs to remove TSO, polyA, and polyG sequences that can negatively impact alignment and mapping rates depending on the sample type and read 2 length.

• TSO sequences are typically observed in <1–5% of reads from Illumina Single-Cell Prep Libraries and are more common in shorter, unfragmented library molecules.
• PolyA sequences occur when reads extend into the polyA tail or reverse-complement polyT capture sequence.
• PolyG sequences arise from reads containing short inserts, which generate dark cycles after processing past the end of the library molecule. Dark cycles are as G bases on Illumina instruments.

Manual pre-trimming is recommended until support for these artifacts is incorporated in a future DRAGEN release (v4.6).

The workflow below produces trimmed FASTQ pairs while trimming only Read 2, allowing the output files to be used directly with the DRAGEN-STAR pipeline.

For more information about using the STAR Mapper with DRAGEN, see the following: https://developer.illumina.com/news-updates/your-pipseq-workflow-is-consolidating-into-dragen

1. Install Cutadapt

# installation
conda create -n cutadapt -c conda-forge -c bioconda cutadapt
# Activate it
conda activate cutadapt
# verify installation
cutadapt --version

2. Set up the input + output paths and run the cutadapt command

r1="/path/to/r1.fastq.gz"
r2="/path/to/r2.fastq.gz"
sample_id="sample_name"

cutadapt \
  --cores 8 \
  --no-indels \
  --pair-filter=any \
  -e 0 \
  -O 12 \
  -U 1 \
  -G AGAGTGAATGGG \
  -G TCAACGCAGAGT \
  -A AAAAAAAAAAAA \
  -A "GGGGGGGGGGGGX;o=6;e=0.15" \
  -m 1:20 \
  -o "${sample_id}_R1.pretrimmed.fastq.gz" \
  -p "${sample_id}_R2.pretrimmed.fastq.gz" \
  "$r1" "$r2"

Parameter Details

Cutadapt option	Purpose
-U 1	Trims 1 base from the 5′ end of Read 2, the transcript read.
-G AGAGTGAATGGG	Removes the TSO sequence from the 5′ side of the transcript read.
-G TCAACGCAGAGT	Removes the SMART PCR sequence from the 5′ side of the transcript read.
-A AAAAAAAAAAAA	Removes poly-A sequence from the 3′ side of the transcript read.
-A GGGGGGGGGGGG o=6;e=0.15"	Removes poly-G sequence from the 3′ side of the transcript read, requiring a 6 bp minimum overlap (o) and allowing 15% errors (e) in the matched region
-O 12	Requires a 12-base adapter overlap, matching the intended 12-base trimming stringency.
-e 0.10	Allows up to 10% mismatch, consistent with DRAGEN adapter trimming behavior.
--no-indels	Disallows insertions/deletions during adapter matching, making the behavior closer to DRAGEN’s mismatch-based adapter trimming.
-m 1:20	Keeps Read 1 effectively unfiltered by length, while requiring the transcript read to be at least 20 nt after trimming.
-p <R2 path>	Path to output trimmed R2 output
-o <R1 path>	Path to output R1 (untrimmed, but matching reads retained from R2)