What Is NGS Data Analysis?

Next-generation sequencing produces large collections of DNA or RNA sequence reads. Bioinformatics methods transform these measurements into structured genomic or transcriptomic results.

NGS data analysis is not a single fixed procedure. Whole-genome sequencing, whole-exome sequencing, targeted DNA sequencing, RNA-seq, metagenomics and single-cell sequencing require different analytical approaches.

This guide focuses primarily on reference-based short-read DNA sequencing and variant discovery. RNA expression, structural variant analysis, de novo assembly and long-read sequencing involve additional or alternative methods.

A typical DNA variant-analysis workflow

1. Obtain raw sequencing reads and associated metadata.
2. Review sequence quality and technical biases.
3. Perform justified preprocessing.
4. Align reads to a compatible reference genome.
5. Assess alignment and coverage metrics.
6. Identify candidate variants.
7. Evaluate variant quality and apply filtering.
8. Annotate findings and interpret the evidence.

Understanding FASTQ Files and Sequencing Reads

FASTQ is a widely used text format that stores sequencing-read identifiers, nucleotide sequences and corresponding per-base quality scores.

A FASTQ record contains four lines: an identifier, the sequence, a separator and a quality string. For commonly used Phred+33 encodings, each quality character represents a numerical Phred quality score.

@example_read_001
ACGTTGCA
+
IIIIHHGF

This is an illustrative record, not experimental data. The sequence and quality string must have the same number of characters.

Phred quality scores

Phred scores are related to the estimated probability of a base-call error: Q = −10 log10(Perror).

These probabilities describe base-call confidence under the quality-score model. They do not guarantee an equal empirical error rate across every experiment or genomic context.

Single-end vs. paired-end reads

Single-end sequencing reads one end of a fragment. Paired-end sequencing generates reads from both ends, providing additional information about their relative orientation and approximate separation. Properly maintained read pairing can improve some mapping and variant-analysis tasks.

FASTQ Quality Control with FastQC and MultiQC

Raw sequencing datasets should be inspected before downstream analysis. Quality-control reports help identify unexpected sequencing characteristics, technical artifacts and potential sample problems.

Common FASTQ quality metrics

FastQC produces individual quality assessments. MultiQC can aggregate reports from multiple tools and samples to help compare results across a project.

Read Trimming, Adapter Removal and Preprocessing

Sequencing-read preprocessing may involve removing library adapters, trimming low-quality regions or excluding reads that do not meet defined criteria.

Adapter sequences may be present when a sequenced fragment is shorter than the sequencing read length. If not handled appropriately, adapters can interfere with downstream alignment.

Common preprocessing operations

• Identifying and removing sequencing adapters.
• Trimming low-quality read ends where justified.
• Filtering reads that fail predefined criteria.
• Maintaining paired-read synchronization.
• Checking quality metrics again after processing.

Excessive trimming can remove informative sequence and reduce the usable length of reads. Processing decisions should follow the data, assay design and requirements of downstream tools.

Tools used for such tasks include Cutadapt and fastp. Their parameters should be documented and selected for the relevant experiment.

Mapping NGS Reads to a Reference Genome

Reference alignment attempts to determine where individual sequencing reads originate within a selected reference genome.

Read mappers use algorithms and indexes designed to efficiently search large reference sequences. The resulting alignments include genomic coordinates, orientation and information about matching or nonmatching positions.

Why is reference selection important?

Read mapping depends on the chosen reference assembly. Different assembly versions can change coordinate systems and relevant reference sequences. Variant calling and annotation must use compatible reference resources.

Mapping quality

Mapping quality is a Phred-scaled indicator of confidence in an alignment placement as reported by the mapper. The exact estimation procedure varies among alignment programs.

Reads from repetitive, duplicated or structurally complex regions may map ambiguously. Such regions require careful interpretation in variant analysis.

SAM, BAM, CRAM and Sequencing Coverage Analysis

Aligned reads are commonly represented using SAM, BAM or CRAM formats. SAM is text-based, BAM is a compressed binary representation of alignment records, and CRAM uses reference-oriented compression techniques.

Common alignment metrics

Depth is not the same as breadth

A region may have high average read depth but contain poorly covered positions. For whole-exome or targeted sequencing, coverage breadth across intended targets may be more informative than a single mean-depth value.

SAMtools supports operations including viewing, sorting, indexing and evaluating alignment data.

Variant Calling: SNPs, Indels and Genotypes

Variant calling identifies candidate differences between sequencing evidence and a reference genome. For short-read DNA analysis, common targets include single-nucleotide variants and short insertions or deletions.

Germline vs. somatic variant calling

Different callers use different statistical models and representations of sequencing evidence. The biological context determines the appropriate analytical strategy.

Short-variant calling does not automatically characterize copy-number changes, repeat expansions or the full range of structural variation. Dedicated methods may be required.

Understanding VCF Files and Variant Filtering

Variant Call Format (VCF) is a standardized format for representing sequence variants, genomic coordinates, alleles and associated metadata. It may also contain genotype and sample-level information.

Core VCF fields

Variant filtering

Not every candidate call has sufficient evidence to retain without qualification. Filtering strategies may evaluate multiple quality annotations, supporting reads, mapping characteristics and study-specific error patterns.

Fixed thresholds are not universally appropriate. Filtering should be calibrated to the calling model, assay characteristics and intended application.

VCF specifications are maintained in the SAMtools/HTS specifications repository .

Variant Annotation and Genomic Interpretation

Variant annotation connects computationally detected genomic changes to gene models, transcripts, sequence consequences, population information and available biological knowledge.

Common annotation categories

• Genomic location: chromosome, position and associated gene features.
• Transcript consequence: effects relative to specific transcripts.
• Protein-level consequence: predicted amino acid changes, where applicable.
• Population context: available frequency information.
• Prior evidence: relevant literature and curated annotations.
• Technical confidence: support from read and variant-quality metrics.

Tools such as Ensembl Variant Effect Predictor can annotate variants relative to selected reference resources. Interpretation requires attention to transcript selection, assembly versions and database releases.

A computational prediction of functional impact is not itself proof of biological effect or clinical significance. Validation and specialist interpretation are required where conclusions have important consequences.

NGS Pipeline Validation, Limitations and Frequently Asked Questions

Reliable NGS analysis requires traceable sample information, compatible reference resources, documented software parameters and suitable quality thresholds.

Important limitations include sequencing errors, coverage gaps, GC-related biases, ambiguous mapping, contamination, reference representation and caller-specific assumptions.

A reproducible analysis should record

• Sample identifiers and sequencing assay characteristics.
• Reference genome assembly and associated files.
• Quality-control summaries and exclusions.
• Software versions and processing parameters.
• Alignment and coverage statistics.
• Variant-calling and filtering methodology.
• Database versions used for annotation.
• Validation strategy and interpretation limits.

Frequently asked questions