Paired-end data

Trim Galore processes paired-end reads in a single pass. R1 and R2 are read, trimmed, and validated together. The per-read order is preserved, which most aligners require, and there are no temp files between Cutadapt invocations like the v0.6.x Perl wrapper used.

Basic invocation

trim_galore --paired sample_R1.fastq.gz sample_R2.fastq.gz

Outputs:

• sample_R1_val_1.fq.gz (validated R1)
• sample_R2_val_2.fq.gz (validated R2)
• sample_R1.fastq.gz_trimming_report.txt (and .json)
• sample_R2.fastq.gz_trimming_report.txt (and .json)

Multiple pairs

--paired accepts any even number of input files and processes them as consecutive pairs in R1, R2, R1, R2, … order:

trim_galore --paired \
    s1_R1.fq.gz s1_R2.fq.gz \
    s2_R1.fq.gz s2_R2.fq.gz \
    s3_R1.fq.gz s3_R2.fq.gz

Don't use *_R1.fq.gz *_R2.fq.gz globs

A glob like *_R1.fq.gz *_R2.fq.gz produces all R1s, then all R2s, not interleaved pairs. Invoke once per sample, or interleave the pairs explicitly as shown above.

Pair validation

Quality and adapter trimming can shorten one read of a pair below the --length cutoff. The pair-validation step runs after per-read trimming and decides what happens to such pairs:

• If both reads pass --length, the pair is written to *_val_1.fq.gz and *_val_2.fq.gz.
• If at least one read is below --length, the pair is discarded by default.

The Read 2 trimming report carries the final pair counts at the end:

RUN STATISTICS FOR INPUT FILE: SLX_R2.fastq.gz
=============================================
1166076593 sequences processed in total

Total number of sequences analysed for the sequence pair length validation: 1166076593

Number of sequence pairs removed because at least one read was shorter than the length cutoff (20 bp): 3357967 (0.29%)

Retaining singletons

When only one of the two reads dropped below the cutoff (e.g. one read had very poor qualities throughout), --retain_unpaired keeps the surviving read in a separate file:

trim_galore --paired --retain_unpaired sample_R1.fq.gz sample_R2.fq.gz

Additional outputs:

• sample_R1_unpaired_1.fq.gz. R1 reads whose mate was too short.
• sample_R2_unpaired_2.fq.gz. R2 reads whose mate was too short.

The per-side cutoff is governed by --length_1 and --length_2 (default 35 bp each).

Per-side trimming

One side of a paired-end library can have biases at a specific end (e.g. RRBS Read 2 has fill-in artifacts at the 5' end). The clip flags remove fixed numbers of bases without touching the other read:

Flag	Effect
--clip_R1 INT	Remove N bases from the 5' end of Read 1.
--clip_R2 INT	Remove N bases from the 5' end of Read 2.
--three_prime_clip_R1 INT	Remove N bases from the 3' end of Read 1 (after quality / adapter trim).
--three_prime_clip_R2 INT	Remove N bases from the 3' end of Read 2.

--rrbs in directional mode auto-sets --clip_R2 2 to mask the 2 bp end-repair bias at the start of Read 2, unless you supply your own --clip_R2. --non_directional skips this auto-clip on purpose.

--discard_untrimmed

--discard_untrimmed keeps only reads where at least one adapter match was found. For paired-end, the pair is discarded only when neither read had an adapter.