cutadapt Documentation

Cutadapt is a tool designed to find and remove unwanted sequences from high-throughput sequencing reads. This includes:

• Adapter sequences: Common in small-RNA sequencing when the read is longer than the molecule.
• Primers: Often found at the start of amplicon reads.
• Poly-A tails: Often removed to clean up RNA sequencing data.

Key features include:

• Error-tolerant searching: Finds sequences even with mismatches.
• Sequence modification: Can modify and filter both single-end and paired-end reads.
• IUPAC support: Adapter sequences can contain IUPAC wildcard characters.
• Demultiplexing: Can be used to demultiplex reads based on barcodes.

When using filtering options (like --minimum-length or --discard-trimmed) on paired-end data, Cutadapt always discards both reads of a pair to keep them in sync. The --pair-filter option determines the logic used to decide if a pair should be discarded:

• --pair-filter=any (default): Discard the pair if at least one read fulfills the criterion.
• --pair-filter=both: Discard the pair only if both reads fulfill the criterion.
• --pair-filter=first: Discard the pair based only on the first read (R1), ignoring the second.

Example behavior with --minimum-length=20:

• any: Pair is discarded if R1 < 20 OR R2 < 20.
• both: Pair is discarded only if R1 < 20 AND R2 < 20.

The quality trimming algorithm (similar to the one used in BWA) removes low-quality bases from the ends of a read.

Process:

1. Subtract the user-provided cutoff from all quality scores.
2. Compute partial sums from the end of the sequence towards the beginning.
3. The sequence is cut at the index where the partial sum is minimal.

This method allows for some high-quality bases to be preserved even if they are surrounded by lower-quality bases, preventing premature trimming of useful data.

When using an index for multiple anchored adapters, Cutadapt checks for 'ambiguous sequences'—sequences that match two or more adapters equally well.

• Default behavior (v5.0+): Cutadapt prints a WARNING and does not trim reads with ambiguous sequences to avoid incorrect assignments.
• With --no-index: Cutadapt will not print the warning and will assign the ambiguous read to the first matching adapter in the list.

Cutadapt can significantly speed up demultiplexing by building an index of adapter sequences instead of checking them one by one. To enable index creation, the following conditions must be met:

1. Anchoring: Adapters must be anchored.
   • 5' adapters: -g ^ADAPTER or -g ^file:adapters.fasta
   • 3' adapters: -a ADAPTER$ or -a file$:adapters.fasta
2. Error Rate: The maximum error rate (-e) must be between 0 and 3.
3. No Wildcards: No IUPAC wildcards can be used, and --match-read-wildcards must not be used.

Optimization Tips:

• Providing --no-indels makes index creation faster and uses less memory.
• You can disable indexing with --no-index.
• If an index is built, you will see Building index of X adapters ... in the output.

The --poly-A trimming algorithm identifies and removes a suffix of the read consisting of a high concentration of A nucleotides.

Algorithm Logic:

1. It evaluates all possible suffixes of the read.
2. It excludes any suffix where non-A nucleotides exceed 20% of the suffix length.
3. For valid suffixes, it calculates a score: +1 for A nucleotides and -2 for non-A nucleotides.
4. The suffix with the highest score is selected for removal. In the event of a tie, the shorter suffix is chosen.

# Conceptual implementation of Poly-A trimming
n = len(s)
best_index = n
best_score = score = errors = 0
for i, nuc in reversed(list(enumerate(range(n)))):
    if nuc == "A":
        score += 1
    else:
        score -= 2
        errors += 1
    if score > best_score and errors <= 0.2 * (n - i):
        best_index = i
        best_score = score

s = s[:best_index]

When processing paired-end reads, Cutadapt validates that the read IDs of R1 and R2 match. If they do not match, Cutadapt prints an error and aborts.

Matching Logic:

• Comments in the FASTQ/FASTA header are ignored.
• If the read ID ends with 1, 2, or 3, these characters are ignored for the comparison.

Examples:

• Properly paired: @my_read/1 a comment and @my_read/2 another comment (the /1 and /2 are ignored).
• Improperly paired: @my_read/1;1 and @my_read/2;1 (the ;1 is considered part of the name and is not ignored, causing a mismatch).

@my_read/1 a comment
@my_read/2 another comment

# This is considered properly paired because /1 and /2 are ignored at the end.

@my_read/1;1
@my_read/2;1

# This is considered improperly paired because ;1 is part of the name.

To prevent erroneously trimming bases due to short random matches, Cutadapt requires a minimum number of bases to align.

• Default: 3 bases.
• Global setting: Use -O or --overlap <value>.
• Adapter-specific setting: Use min_overlap=<value> within the adapter string.
• Constraint: The minimum overlap cannot be set for anchored adapters, as they require full-length matches by definition.
• Linked adapters: The minimum overlap is applied separately to the 5' and 3' components of a linked adapter.

cutadapt -O 5 -a "ADAPTER;min_overlap=10"

Cutadapt applies modifications to each read in a specific, deterministic order. Steps not requested via command-line options are skipped. The sequence is:

1. Unconditional base removal: --cut
2. Quality trimming: -q
3. Adapter trimming: -a, -b, -g (and uppercase versions)
4. Poly-A/poly-T trimming: --poly-a
5. Read shortening: --length
6. N-end trimming: --trim-n
7. Length tag modification: --length-tag
8. Read name suffix removal: --strip-suffix
9. Name prefix/suffix addition: -x/--prefix and -y/--suffix
10. Read renaming: --rename
11. Zero capping: Replacing negative quality values with zero