Adaptor Sequence Illumina: A Guide for Trimming & Design

When a sequencing run finishes, the first quality plots of the FASTQ files often show overrepresented sequence at the 3′ end. That usually means the insert was shorter than the read, and the sequencer continued into the adapter.

At that point the Illumina adaptor sequence becomes an analysis problem and stops being a matter of looking up a reference sheet. Trimming the wrong sequence, trimming in the wrong orientation or ignoring the library structure can distort alignments. It can also inflate soft clipping, mis-handle low-complexity libraries and make downstream QC harder than it needs to be.

For routine whole-genome work, the result may be noisier alignments. For amplicons, CRISPR validation, construct verification, targeted panels or synthetic biology libraries, it can directly affect whether the designed sequence can be trusted. Adapter handling sits at the boundary between wet-lab prep and computational interpretation, and it deserves more attention than a pasted list of oligos.

Why Illumina Adapter Sequences Matter

Illumina made adapter sequences a formal standard reference in October 2015 with the Illumina Adapter Sequences Document (Document #1000000002694 v00), which gave labs an authoritative source for the exact oligos used across kits such as TruSeq and Nextera (Illumina Adapter Sequences Document). The release mattered because adapter trimming only works when the software searches for the sequence present in the library.

The practical trigger is simple. When read lengths exceed insert sizes, reads run through the insert and into adapter sequence. Illumina platforms such as NovaSeq can produce paired-end reads up to 300 bp from the documented platform context in that same reference, so read-through is common in short-insert libraries and aggressively sized fragments.

What goes wrong when adapters remain

Untrimmed adapters affect the analysis in several ways.

• Aligners see non-genomic sequence and either soft-clip it or force poorer placements.
• Variant calling gets noisier because mismatched read ends can create artifactual evidence.
• Quantification pipelines inherit junk sequence that never belonged to the biological insert.
• Assembly and construct validation become more fragile when read ends carry synthetic sequence rather than template-derived sequence.

For synthetic biology work, computational errors can carry into experimental interpretation. A bad trim can look like a failed build, a strange junction or a low-frequency edit that does not exist.

Practical rule: If the read architecture is unclear before trimming, stop and reconstruct the library design first. Sequence cleanup should follow library structure, and a guess about that structure is a poor substitute for reconstructing it.

Why this matters beyond basic QC

Many teams treat adapter trimming as a checkbox in a pipeline. Adapter sequence choice affects demultiplexing assumptions, trimming sensitivity and what counts as a usable read after cleanup. It also changes how poor runs are interpreted. High adapter content may reflect short inserts, a prep issue or a library class where read-through is expected.

The useful question concerns the exact adapters and their orientation. It asks which read carries them, which kit they came from and what their abundance says about the library.

The Anatomy of an Illumina Adapter

An Illumina adapter is a functional assembly of sequence elements that let a DNA fragment bind the flow cell, accept sequencing primers and carry sample identity through indexing.

The structural pieces that matter

Illumina adapter architectures include P5 and P7 sequences for flow cell binding, Read 1 and Read 2 sequencing primer regions, and index sequences for multiplexing. These are the pieces to understand before troubleshooting anything.

The table below gives the working model of the components.

Component	What it does	Why you care in practice
P5 / P7	Bind library molecules to complementary oligos on the flow cell	If these are wrong or incompatible, cluster generation fails or behaves poorly
Read 1 / Read 2 primer binding sites	Provide annealing sites for sequencing primers	These define where sequence acquisition begins and how read orientation is interpreted
i7 and i5 indexes	Encode sample identity for pooled libraries	Wrong index handling causes demultiplexing loss, cross-sample confusion, or both
Adapter junctions	Connect the library insert to sequencing-compatible structure	These are the exact regions trimming tools look for in read-through events

P5 and P7 are not trimming targets in the abstract

People often talk about “the adapter” as if there were one universal sequence to remove. Trimming usually targets the sequence encountered after reading through the insert into the adapter-derived region. That read-through sequence is related to the broader adapter design, and the trimming target depends on kit architecture and read orientation.

P5 and P7 are central to the molecule, yet the sequence typed into Cutadapt often differs from them. The exact trimming sequence has to match the sequence expected at the read end.

Indexes are operational

Indexes are often treated as metadata, but they are physical sequence components embedded in the library. Illumina’s documented adapter scheme includes Index 1 (i7) and Index 2 (i5) components, and for many workflows that distinction matters during both demultiplexing and troubleshooting.

A practical consequence is that adapter identity and index identity cannot be separated cleanly in real runs. If a sample sheet is wrong, demultiplexing accuracy may suffer and the interpretation of which sequence should appear where may also go wrong.

Much adapter troubleshooting is library-structure troubleshooting that presents itself as a bioinformatics problem.

The annealing detail people forget

In the formal adapter documentation, Illumina notes that for certain indexed adapter components, only the last 12 nucleotides are complementary for annealing. That detail helps explain both why the forked adapter structure works and why dimers are such a persistent wet-lab problem in low-input prep. Those short complementary regions are enough to create the intended ligation-ready structure. They are also enough to create unwanted adapter-adapter products when the insert population is weak.

The design works well when the library is healthy and becomes costly when the insert population is weak.

Quick Reference for Common Adapter Sequences

Most readers who look for Illumina adaptor sequences need the trimming sequence immediately. The table below is the fast lookup.

The safest habit is still to confirm the exact kit and document version before production analysis. For day-to-day trimming, a short reference table covers most of what comes up in standard workflows.

Common Illumina adapter trimming sequences

Kit Family	Universal Adapter Sequence to Trim
TruSeq	AGATCGGAAGAGCACACGTCTGAACTCCAGTCA
Nextera	AGATCGGAAGAGCACACGTCTGAACTCCAGTCA
Illumina DNA Prep / Illumina RNA Prep / PCR kits	AGATCGGAAGAGCACACGTCTGAACTCCAGTCA
Nextera Mate Pair	CTGTCTCTTATACACATCT+AGATGTGTATAAGAGACAG

How to use this table correctly

The library context decides which sequence applies, because the string alone gives too little information.

• For many standard Illumina kits, the universal trimming sequence is the same.
• For mate-pair libraries, the composite sequence matters because each adapter half may need to be assessed independently.
• For indexed libraries, index sequence handling and generic adapter trimming are related and have to be handled separately.

A few lookup habits that prevent mistakes

• Check the kit family first. The phrase “Illumina library” is too general.
• Confirm whether the library is paired-end or single-end before choosing tool settings.
• Look at overrepresented sequences in raw QC and compare them to expected adapter-derived sequence.
• Do not recycle an old adapters.fa file across unrelated kits.

If the run comes from a shared core, ask for the exact prep kit name. Approximations such as “TruSeq-like” or “Nextera-based” create trimming errors faster than anticipated.

Understanding Major Illumina Adapter Variants

Illumina-compatible adapters solve different problems. Some are built around ligation, some come from transposase-based fragmentation and tagging, and others differ mainly in indexing strategy. The indexing difference matters most in a large pooled run, where it can leak reads across samples.

TruSeq and ligation-based workflows

TruSeq-style libraries are the classic ligation-based model. The workflow fragments the DNA, repairs the ends, adds A-tails where appropriate and ligates adapters. That architecture is predictable, which is why it is still easy to reason about during troubleshooting.

The upside is clarity. The insert boundaries are usually known, and adapter interpretation in reads is relatively straightforward. The downside is that low-input work can be sensitive to dimer formation and ligation inefficiency.

Nextera and transposase-based workflows

Nextera libraries follow a different design. Tagmentation combines fragmentation and adapter addition in one step, which changes where errors enter the workflow.

The workflow gains speed and convenience, and it also brings kit-specific sequence expectations that people mis-handle when they assume every Illumina library behaves like TruSeq. That assumption is one reason trimming files get copied between projects without enough review.

CDI versus UDI is not a minor upgrade

Indexing strategy is where many pooled experiments become fragile. Combinatorial dual indexing, or CDI, reuses i5 and i7 indexes in combinations. Unique dual indexing, or UDI, assigns a fully unique pair to each sample.

That distinction matters most on high-throughput patterned flow cells. Illumina’s adapter portfolio bulletin states that UDI kits reduce index hopping to less than 0.1% on patterned flow cells like NovaSeq.

When UDI is the right choice

For small, simple pools, CDI may still be operationally acceptable. UDI is usually the better engineering decision for high-plex designs and sensitive rare-variant work. The same holds for synthetic biology libraries where sample identity is tightly coupled to construct identity.

The practical summary follows.

• Choose CDI when the pool is modest, the assay is tolerant, and index reuse won’t compromise interpretation.
• Choose UDI when contamination by reassigned reads would create false positives or sample ambiguity.
• Choose UDI by default when going beyond 96-plex, which the same Illumina bulletin identifies as a critical context for high-plex synthetic biology and related workflows.

The value of UDI depends on a narrower question about the assay, which is what one misassigned read would mean for it.

The hidden trade-off

UDI improves specificity and also raises operational demands. Teams have to manage index inventory carefully, verify sample sheets and maintain discipline across automation steps. That work is worthwhile and it carries a cost. Many sequencing failures blamed on the instrument are failures in index bookkeeping.

How to Detect and Trim Adapters from FASTQ Files

Most trimming failures arise when the operator does not know which adapter is present, which read contains it or how aggressively the trimming step should prune read-through. The software itself is rarely the cause.

For many Illumina kits, the universal adapter trimming sequence is AGATCGGAAGAGCACACGTCTGAACTCCAGTCA. For Trimmomatic a commonly used parameter string is ILLUMINACLIP:adapters.fa:2:30:10.

Start by confirming the library structure

Before running anything, confirm the following.

• Identify the prep kit from the wet-lab record.
• Check whether the run is single-end or paired-end.
• Inspect raw QC for overrepresented sequence and read-end contamination.
• Decide whether adapter read-through is expected because of short inserts.

If the prep team cannot say what the library is, ask for the sample sheet, index layout and prep protocol. That usually resolves the ambiguity faster than trying six trimming settings.

For a broader prep context, the details of library construction in this NGS library prep overview are a useful operational reference.

Cutadapt examples

Cutadapt is the cleanest tool when explicit control is needed.

Single-end Cutadapt

cutadapt \
  -a AGATCGGAAGAGCACACGTCTGAACTCCAGTCA \
  -m 20 \
  -o sample.trimmed.fastq.gz \
  sample.fastq.gz

These choices have the following effects.

• -a specifies the 3′ adapter to trim.
• -m 20 discards reads that become too short to be useful after trimming.
• Keeping the command minimal is often better than over-tuning on the first pass.

Paired-end Cutadapt

cutadapt \
  -a AGATCGGAAGAGCACACGTCTGAACTCCAGTCA \
  -A AGATCGGAAGAGCACACGTCTGAACTCCAGTCA \
  -m 20 \
  -o sample_R1.trimmed.fastq.gz \
  -p sample_R2.trimmed.fastq.gz \
  sample_R1.fastq.gz sample_R2.fastq.gz

This is the right baseline when both mates can run through short inserts into adapter sequence.

Trimmomatic examples

Trimmomatic is common in established pipelines because it integrates adapter clipping with additional cleanup steps.

Prepare an adapters file

An adapters.fa file contains the sequence to be trimmed, as in this example.

>IlluminaUniversal
AGATCGGAAGAGCACACGTCTGAACTCCAGTCA

Single-end Trimmomatic

trimmomatic SE -phred33 \
  sample.fastq.gz sample.trimmed.fastq.gz \
  ILLUMINACLIP:adapters.fa:2:30:10 \
  LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:20

Paired-end Trimmomatic

trimmomatic PE -phred33 \
  sample_R1.fastq.gz sample_R2.fastq.gz \
  sample_R1.paired.fastq.gz sample_R1.unpaired.fastq.gz \
  sample_R2.paired.fastq.gz sample_R2.unpaired.fastq.gz \
  ILLUMINACLIP:adapters.fa:2:30:10 \
  LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:20

The parameter string ILLUMINACLIP:adapters.fa:2:30:10 has three numeric fields.

• 2 is the seed mismatch setting.
• 30 is the palindrome clip threshold.
• 10 is the simple clip threshold.

Those values are widely reused because they strike a practical balance. They catch real adapter sequence without turning every low-quality tail into a false adapter match.

fastp examples

fastp is convenient when a compact command and built-in reporting are wanted.

Single-end fastp

fastp \
  -i sample.fastq.gz \
  -o sample.trimmed.fastq.gz \
  --detect_adapter_for_pe \
  --length_required 20

For single-end data, explicit adapter specification is still preferable when the library is known. Auto-detection is useful, and it should still be combined with knowledge of the library.

Paired-end fastp

fastp \
  -i sample_R1.fastq.gz \
  -I sample_R2.fastq.gz \
  -o sample_R1.trimmed.fastq.gz \
  -O sample_R2.trimmed.fastq.gz \
  --detect_adapter_for_pe \
  --length_required 20

The video below is a visual walkthrough of trimming logic in practice.

Special case for Nextera Mate Pair

Nextera Mate Pair libraries need extra care. The documented composite trimming sequence is CTGTCTCTTATACACATCT+AGATGTGTATAAGAGACAG, which supports independent assessment of each adapter half in mate-pair cleanup. Treating mate-pair data like ordinary short-insert paired-end data often keeps the wrong reads and discards the ones that are needed.

A universal Illumina trimming command copied into mate-pair analysis gives no assurance that the right reads are kept.

Interpreting Adapter Trimming Reports and Metrics

Trimming continues after the command exits cleanly, because the report shows what the adapter burden says about the library.

What trimming reports tell you

Most tools report a few core signals:

• How many reads contained adapter sequence
• How many bases were removed
• How many reads became too short and were discarded
• Whether contamination was symmetric across R1 and R2

Those values show whether the run had a normal short-insert read-through pattern, a prep problem or a configuration mismatch.

How to read asymmetry

If one mate trims heavily and the other barely trims, the cause is unlikely to be biology. The adapter specification may be incomplete, one read may have lower quality, or the library structure may cause one end to encounter adapter more often.

That kind of asymmetry often points to a process issue and rarely to a feature of the sample.

BCL Convert metrics are especially useful

Illumina’s Adapter_Metrics.csv in BCL Convert v1.10+ reports precise counts of adapter bases per lane and sample. Illumina states that these metrics support computation of true genomic yield. It connects this practice to preventing alignment bias (BCL Convert Adapter_Metrics documentation).

That matters because total sequenced bases differ from biologically informative bases. If adapter content is high, the apparent depth can be misleading.

A good companion for thinking about usable depth is this overview of DNA sequencing coverage, especially when raw output has to be distinguished from effective coverage.

What to do with the numbers

The trimming report and the BCL Convert metrics are read together.

Signal	Likely interpretation	Follow-up
Low adapter trimming	Inserts were generally longer than read length	Proceed, but still review read quality and duplication
Moderate adapter trimming	Typical short-insert read-through	Usually acceptable if expected for the assay
High adapter bases for one sample only	Sample-specific prep issue or size-selection drift	Check fragment analysis and pooling notes
High adapter burden across many samples	Run-wide insert sizing or prep design issue	Review library prep and sequencing read length choice

High adapter content can be an accurate warning that the library was built shorter than the sequencing strategy assumed. Adapter content alone does not establish a trimming failure.

Troubleshooting Common Adapter-Related Sequencing Issues

When a run looks bad, adapter handling is often where the postmortem starts. The hard part is separating computational symptoms from wet-lab causes.

Adapter dimers are small, common, and expensive

Adapter dimers form because the forked Y-adapter design contains complementary 12-nucleotide regions. In low-concentration libraries, including single-cell and CRISPR sgRNA work, dimers can consume sequencing reads and lower yield. The same Illumina knowledge reference also notes mitigation approaches such as optimized SPRI bead ratios like 1.8x and alternative designs such as NEBNext stem-loop structures (Illumina knowledge reference on library preparation and dimers).

This is one of the most consequential adapter-related facts to keep in mind. Dimers waste reads and also crowd out the library that was intended for sequencing.

How dimers show up in data

Common signs include:

• Very high abundance of short reads after trimming
• Overrepresented pure adapter sequence in raw QC
• Unexpectedly poor usable yield despite nominal sequencing output
• Libraries that quantify weakly but still amplify strongly

The last pattern is frequently overlooked. Dimers amplify efficiently because they are short, so a library can look PCR-productive while being analytically poor.

What works against dimers

The most effective interventions are upstream of sequencing.

• Tighten bead cleanup conditions when the insert distribution allows it.
• Avoid pushing low-input ligations past their comfortable range.
• Re-check adapter:insert balance instead of only increasing PCR.
• Consider alternative adapter chemistries if the assay class repeatedly produces dimers.

Aggressive read trimming cannot repair a dimer-heavy library on its own, because the instrument has already spent the reads by then.

Index hopping needs a different response

Dimers and hopping are often grouped together because both create contamination. They are separate problems.

Index hopping is a sample-assignment problem. If the assay is sensitive to rare events, the recommended answer is usually to move to unique dual indexing rather than trying to rescue interpretation later with stricter downstream thresholds.

Demultiplexing failures usually start with metadata

When demultiplexing collapses, check these first:

1. Was the sample sheet correct for index orientation and kit type?
2. Were i5 expectations instrument-compatible?
3. Were the oligos used the ones assumed by the sequencing setup?
4. Did pooling records drift from plate layout?

Most demultiplexing investigations become much shorter once someone compares the wet-lab worksheet to the sequencer configuration line by line.

When a run fails to demultiplex cleanly, the investigation begins with the index map and the exact kit. BCL conversion is a later suspect.

Guidelines for Designing Custom Illumina Adapters

Custom adapter design combines bioinformatics, oligo chemistry and assay engineering. It is a systems problem, and the biggest mistake is to treat it as a sequence-formatting exercise.

Preserve the parts Illumina chemistry expects

Any custom design has to remain compatible with the sequencing ecosystem. That means preserving the required functional logic around flow-cell binding, primer annealing and indexing, even when assay-specific features such as custom barcodes or UMIs are added.

If a custom motif disrupts the expected architecture, the failure may show up as poor clustering, bad indexing behavior or strange read structures rather than an obvious design error.

Design for analysis as well as synthesis

A custom adapter should be easy to reason about computationally.

Good design choices include:

• Distinct barcode space that won’t collapse under sequencing error
• Clear positional logic for where barcode, UMI, and insert begin
• Minimal ambiguity at junctions so trimming and parsing remain deterministic
• Avoidance of problematic secondary structures that complicate ligation or amplification

A parser that needs a page of exception handling points to a problem in the adapter design.

Validate with a small pilot first

Before scaling a custom scheme, test whether the resulting reads are easy to demultiplex, trim and map. That is especially important when custom constructs will feed into genome-scale or combinatorial workflows.

A useful operational reference for how these design choices interact with broader library construction is this overview of genomic DNA libraries.

QC for custom adapters should be explicit

Amplification alone is weak evidence that the design works.

Use a validation checklist:

• Confirm oligo identity and purity from the synthesis provider.
• Run a pilot library before committing a full plate or production batch.
• Inspect raw read structure manually in a subset of FASTQs.
• Verify that trimming behavior matches the intended junctions.
• Check whether off-target short products dominate amplification.

The best custom adapters are easy for humans and software to interpret under failure conditions. Compatibility with Illumina is a minimum requirement.

Frequently Asked Questions About Illumina Adapters

What happens if I use the wrong adapter sequence in trimming software

Usually one of two things happens. The software fails to trim real contamination, or it trims incompletely and leaves adapter-derived sequence behind. In harder cases, it clips reads at the wrong places and reduces usable sequence unnecessarily.

Can I mix adapters from different Illumina kits in one sequencing pool

Safe pooling cannot be assumed. Compatibility depends on the full library architecture, indexing scheme and sequencing setup. If kits differ in index design, orientation expectations or read structure, pooling them can create demultiplexing and analysis problems.

How short does the insert need to be before trimming becomes essential

Trim as soon as read-through into adapter sequence is occurring. In practice, if the sequenced read length can exceed the insert, trimming is required.

Does adapter orientation matter in paired-end sequencing

Yes. Orientation matters in both trimming and demultiplexing. A correct sequence in the wrong orientation can still produce wrong results.

Should I trust auto-detection of adapters

Only if it agrees with the known library structure. Auto-detection is convenient for confirmation and cannot replace knowing what was ligated.

If trimming removes many reads, is the trimmer too aggressive

The trimmer is not necessarily too aggressive. Heavy read loss may correctly report short inserts, adapter dimers or poor library composition. The right response is to inspect the library assumptions, since relaxing the trimming settings alone would leave the underlying cause in place.