Illumina Sequencing
Illumina sequencing is a short read, high throughput DNA sequencing technology based on sequencing by synthesis (SBS). It works by detecting fluorescently labeled nucleotides as they are incorporated into a growing DNA strand, using reversible terminators that add one base at a time. This guide is for molecular biologists, bioinformaticians, and clinical lab scientists who need a practical, source bounded understanding of Illumina sequencing from library preparation through data interpretation. The method is widely used for whole genome sequencing, transcriptomics, epigenomics, and targeted panels because of its accuracy, scalability, and cost effectiveness. For an authoritative technical reference on the underlying chemistry, see the NCBI Bookshelf introduction to next generation sequencing.
The workflow involves fragmenting DNA, attaching adapters, amplifying fragments on a solid surface, and imaging each cycle of base incorporation. The base calls are processed into reads typically 75 to 300 base pairs long. Although Illumina dominates the short read market, the user must understand the technology's systematic biases and quality metrics to avoid misinterpretation. The EMBL EBI Training resource on sequencing technologies provides an excellent overview of where Illumina fits among competing platforms.
At a Glance
| Aspect | Detail |
|---|---|
| Technology basis | Sequencing by synthesis with reversible terminators |
| Read length | 75 to 300 base pairs (typical) |
| Output per run | 100 Mb to 6 Tb depending on instrument |
| Error type | Substitutions dominate, insertion deletions are rare |
| Common applications | Whole genome, exome, RNA seq, ChIP seq, metagenomics |
| Scalability | from individual gene panels to population scale |
| Key limitation | Short reads limit de novo assembly and structural variant detection |
Decision Criteria for Choosing Illumina
Not every sequencing question is best answered by Illumina. The following decision points help you assess whether Illumina is appropriate for your project.
Read length requirement. If your goal requires resolving repetitive regions, full length transcript isoforms, or long structural variants, a short read platform may underperform. Illumina reads cannot span entire transposable elements or highly repetitive centromeres. For such cases, consider long read technologies. However, for variant detection in unique regions, Illumina's high base quality (often Q30 or higher) is superior.
Depth of coverage and cost. Illumina offers the lowest per base cost among accurate sequencing platforms. For applications like tumor normal paired sequencing where 100x coverage is needed, Illumina is the standard. The Galaxy Training Network workflow for variant calling shows how to plan coverage depth based on sensitivity requirements.
Sample throughput. If you need to sequence hundreds of samples simultaneously, Illumina's multiplexing capability (using index adapters) allows pooling libraries on a single lane or flow cell. This is ideal for large cohort studies but may be wasteful for a single sample where a smaller platform would suffice.
Error tolerance. Illumina substitution error rates are below 0.1% for most instruments, making it adequate for germline variant detection. For low frequency somatic mutations (below 1% allele fraction), specialized error correction or unique molecular identifiers are required. The Bioconductor documentation for variant tools provides packages that model sequencing errors and improve detection.
Practical Workflow: From Library to Base Calls
The standard Illumina workflow consists of four major stages: library preparation, cluster generation, sequencing by synthesis, and data output. Each stage requires careful attention to detail.
Step 1: Library preparation. DNA or RNA is fragmented to a target size (often 200 to 500 base pairs). End repair, A tailing, and adapter ligation are performed. Adapters contain sequences for flow cell binding, sequencing primer binding, and sample specific indexes. For RNA, reverse transcription and strand specific steps are needed. After ligation, the library is size selected and PCR amplified. Over amplification can introduce duplicates and bias. The NCBI Sequence Read Archive best practices recommend documenting library metrics such as fragment size distribution and concentration.
Step 2: Cluster generation. Libraries are denatured and loaded onto a flow cell coated with oligonucleotides complementary to the adapters. Each molecule is bridge amplified to form a clonal cluster of about 1,000 identical copies. This step amplifies the signal and ensures that each cluster produces a distinct fluorescence intensity per cycle. In newer patterned flow cells, clusters are generated on nanowells, which reduces crowding and improves base call quality.
Step 3: Sequencing by synthesis. All four labeled nucleotides are added simultaneously. Each nucleotide blocks further incorporation after the first addition. The flow cell is imaged to record which base was added to each cluster. The terminator is then chemically cleaved, and the cycle repeats. Typically, read 1 is sequenced first, then an index read, then read 2 for paired end runs. The number of cycles determines read length.
Step 4: Data output and base calling. Images are processed to produce intensity files, then converted to base calls and quality scores (Phred scores). The output is in FASTQ format, which contains the sequence and per base quality. The Galaxy Training Network introduction to FASTQ explains how to interpret these quality scores and filter reads.
Quality Checks and Data Validation
Raw data from Illumina instruments must be validated before downstream analysis. The following checks are standard.
Per base sequence quality. Use tools like FastQC to examine the quality score distribution across read positions. A typical Illumina run shows declining quality toward the 3 prime end. If median quality drops below Q20 (1% error) in the last 20 bases, consider trimming or using a paired end conservative approach.
Duplication levels. PCR duplicates appear as identical read pairs mapping to the same genomic coordinates. High duplication rates (above 20%) indicate library amplification bias or low input DNA. The NCBI Bookshelf guide on sequence preprocessing recommends marking duplicates with tools like Picard MarkDuplicates before variant calling.
Adapter contamination. Adapter sequences often appear when fragment length is shorter than read length. Use a trimming tool (e.g., Trimmomatic, Cutadapt) to remove adapters. The EMBL EBI Training material on quality control provides command line examples.
GC bias. Illumina libraries with extreme GC content (below 20% or above 80%) are underrepresented. If your target region is GC rich, consult the Bioconductor documentation on GC bias correction to avoid false negative calls.
Common Mistakes and How to Avoid Them
Even experienced groups make avoidable errors in Illumina sequencing. Here are the most frequent pitfalls.
Mistake 1: Over amplification of libraries. PCR cycles beyond 12 introduce duplicate reads and bias allele frequencies. Use a low cycle number and quantify libraries accurately with qPCR or fluorometry. The NCBI Sequence Read Archive submission guidelines emphasize reporting the number of PCR cycles in metadata.
Mistake 2: Ignoring index cross talk. Multiplexing with index adapters can cause misassignment of reads to the wrong sample (index hopping). This is more common on patterned flow cells. Using unique dual indexes instead of combinatorial indexing reduces this error.
Mistake 3: Assuming uniform coverage. Not all genomic regions sequence equally. Repetitive, high GC, or extremely low GC regions are systematically undercovered. In a recent study on an icefish genome assembly, the authors noted gaps in repeat rich regions and recommended using Chromosome scale genome assembly methods that leverage multiple sequencing technologies.
Mistake 4: Neglecting to trim adapters and low quality bases. Adapter contamination leads to false alignments and variant calls. Always run a quality control pipeline before analysis. The Galaxy Training Network offers a complete QC workflow that adapts to different library types.
Mistake 5: Using default parameters without validation. Variant callers, mapper settings, and quality filters should be tuned for your specific organism and read length. For example, avian influenza genome surveillance requires different mapping parameters than human genome sequencing, as shown in a study of Highly Pathogenic Avian Influenza A(H5N1) clade 2.3.4.4b that used optimized variant detection.
Limits of Interpretation and Uncertainty
Illumina sequencing provides high accuracy at the single base level, but the resulting data have inherent limits.
Short read assembly limitations. De novo genome assembly from short reads alone produces fragmented contigs. Repetitive regions longer than the read length collapse or become misassembled. A recent tool called Stash was developed to detect misassemblies by using stochastic tile hashing, highlighting the persistent problem of genome misassembly in short read data.
Structural variant detection. Indels larger than 50 base pairs are difficult to detect with 150 bp reads. Paired end mapping can infer larger deletions and inversions, but breakpoints are often uncertain. For applications like non invasive prenatal testing, methods rely on allele counting rather than direct structural variant detection.
Transcript isoform resolution. Short reads cannot distinguish between alternatively spliced isoforms that share the same few exons. Long read RNA sequencing is superior for full length isoform discovery. A benchmark comparison in 2024 showed that nanopore long reads captured more isoforms than Illumina short reads for single cell and spatial RNA sequencing.
Methylation analysis. While bisulfite converted Illumina libraries can map methylation, the conversion step damages DNA and reduces coverage. The interpretation of methylation levels is also complicated by incomplete bisulfite conversion and non random variation.
Clinical variant interpretation. A high quality Illumina call still requires orthogonal validation (e.g., Sanger sequencing) for rare or novel variants, especially when they affect clinical decisions. The false positive rate at low coverage or in homopolymer runs can be significant.
Frequently Asked Questions
What is the typical error rate for Illumina sequencing? The substitution error rate is approximately 0.1% for standard runs, but this varies by instrument, read position, and sequence context. Quality scores (Phred) quantify the probability of error. For example, a Q30 score means a 1 in 1,000 error probability. The error rate is lower for the first 100 bases and increases toward the end of the read.
Can Illumina sequence long repetitive regions? No. Illumina reads are too short to span most repetitive elements. Repetitive sequences cause mapping ambiguity and assembly collapse. For such regions, long read technologies (PacBio or Oxford Nanopore) are recommended, sometimes combined with Illumina for polishing.
How do I choose the appropriate read length and coverage for my project? For human whole genome sequencing, 150 bp paired end reads at 30x coverage are standard for variant detection. For microbial genomes, 75 bp reads at 50x may suffice. For RNA seq, 75 to 100 bp single or paired end reads are common. Deeper coverage improves sensitivity for low frequency variants but increases cost.
What causes index hopping and how can I prevent it? Index hopping occurs when free index primers from one sample ligate to a library fragment from another sample during cluster generation, leading to reads assigned to the wrong sample. Using unique dual indexes (two different index sequences per sample) dramatically reduces cross sample contamination. This is now standard practice for multiplex Illumina runs.
References and Further Reading
- NCBI Bookshelf introduction to next generation sequencing , a free textbook covering sequencing chemistry and data types.
- EMBL EBI Training on sequencing technologies , official training modules with interactive exercises.
- Galaxy Training Network workflow for variant calling , practical, reproducible pipelines for Illumina data analysis.
- Bioconductor documentation for sequencing quality control , open source software packages and vignettes.
- NCBI Sequence Read Archive submission guidelines , metadata requirements and best practices for data deposition.
- Chromosome scale genome assembly and annotation of Jonah's icefish , a case study using Illumina plus long reads.
- Rapid Expansion of Highly Pathogenic Avian Influenza A(H5N1) clade 2.3.4.4b , application of Illumina sequencing in viral surveillance.
- Genome misassembly detection using Stash , addressing short read assembly errors.
- Benchmark of bioinformatics methods for single cell and spatial RNA seq nanopore long reads , comparison of short and long read sequencing performance.
- Allele counting based non invasive prenatal paternity testing for trisomy 21 , an example of clinical Illumina application.