- Exon, intron, untranslated region (UTR), and coding sequence (CDS) annotation is an important part of structural genome annotation. It identifies and describes the functional regions within genes and transcripts, providing information about how genes are organized, how RNA molecules are processed, and which parts of a transcript encode proteins. Accurate annotation of these regions is essential for genome analysis, gene expression studies, alternative splicing research, comparative genomics, variant interpretation, and functional genomics.
- A typical eukaryotic protein-coding gene may contain several exons separated by introns. When a gene is transcribed, its initial RNA product, known as pre-mRNA, generally contains both exon-derived and intron sequences. During RNA processing, introns are removed through splicing and exons are joined to form mature messenger RNA (mRNA). However, the term exon does not mean that every nucleotide in an exon encodes a protein. Exons may contain coding sequence, 5′ untranslated regions, 3′ untranslated regions, or combinations of these elements. The coding sequence is the portion of a protein-coding transcript that is translated into amino acids, whereas UTRs are transcribed regions that remain untranslated.
- Exons are genomic regions retained in a mature RNA molecule after RNA processing. In protein-coding genes, exons may contain coding regions and untranslated regions. The first exon of a transcript may contain part of the 5′ UTR followed by the beginning of the coding sequence, while the final exon may contain the end of the coding sequence followed by the 3′ UTR. Some exons may consist entirely of untranslated sequence. Exon boundaries are therefore important for identifying transcript structure and determining how different transcript isoforms are produced.
- Exon annotation involves identifying the genomic coordinates of each exon, assigning exon numbers within a transcript, determining their orientation, and linking them to the corresponding gene and transcript identifiers. Exon coordinates are recorded according to the reference genome and its coordinate system. For a gene on the positive strand, exons generally appear in increasing genomic coordinate order in the direction of transcription. For a gene on the negative strand, the transcriptional order is reversed. Annotation software must account for strand orientation when reconstructing transcripts and assigning exon positions.
- Introns are intervening regions of a gene that are transcribed into pre-mRNA but are generally removed during RNA splicing. They separate exons in many eukaryotic genes and can vary considerably in length, sequence composition, and number. Introns often contain conserved splice-site signals that guide the RNA-splicing machinery. The most common spliceosomal introns follow the GT–AG rule, in which the intron typically begins with GT in genomic DNA and ends with AG. In the corresponding RNA, these positions are GU and AG. Other introns, including those processed by the minor spliceosome, may have different boundary signals.
- Intron annotation identifies the genomic intervals between consecutive exons within a transcript. It also records splice donor and acceptor sites, intron length, strand orientation, and relationships with alternative transcripts. Intron boundaries are particularly important because incorrect splice-site annotation can change the predicted coding sequence, disrupt reading frames, or produce incorrect protein sequences. Introns may also contain regulatory elements, repetitive sequences, noncoding RNA genes, and sequences involved in alternative splicing or transcriptional regulation. Consequently, introns should not be considered biologically inactive simply because they are removed from mature mRNA.
- The 5′ untranslated region, or 5′ UTR, is the portion of a mature mRNA located upstream of the translation start site. It is transcribed into RNA but is not translated into the protein. The 5′ UTR can influence translation initiation, ribosome recruitment, mRNA stability, and the regulation of protein production. It may contain upstream open reading frames, RNA secondary structures, internal regulatory sequences, and binding sites for RNA-binding proteins or regulatory molecules. Changes in 5′ UTR sequence can therefore affect gene expression even when the protein-coding sequence remains unchanged.
- The 3′ untranslated region, or 3′ UTR, is the region of a mature mRNA located downstream of the translation termination codon. Like the 5′ UTR, it is transcribed but not translated into the protein. The 3′ UTR contributes to the regulation of mRNA stability, localization, degradation, and translation efficiency. It may contain binding sites for microRNAs, RNA-binding proteins, and other regulatory factors. It also commonly contains signals involved in cleavage and polyadenylation, which determine the formation of the mature 3′ end of the transcript. Alternative polyadenylation can produce transcripts with different 3′ UTR lengths, potentially changing their regulatory properties.
- UTR annotation requires the identification of the translation start and stop positions within each protein-coding transcript. The 5′ UTR is located between the transcription start site and the translation start site, while the 3′ UTR extends from the translation termination site to the transcript’s annotated 3′ end. These regions can differ among transcript isoforms of the same gene because alternative promoters, alternative splicing, and alternative polyadenylation can generate transcripts with different UTR structures. Accurate UTR annotation is therefore important for interpreting gene regulation and transcript diversity.
- The coding sequence, commonly abbreviated as CDS, is the portion of a protein-coding transcript that is translated into a polypeptide. It begins at a translation initiation codon, usually ATG in DNA or AUG in mRNA, and continues through the codons that specify amino acids until a translation termination codon is reached. The standard genetic code contains three stop codons: TAA, TAG, and TGA in DNA, corresponding to UAA, UAG, and UGA in RNA. Stop codons signal termination of translation and do not encode amino acids.
- In annotation files, the CDS is usually represented as one or more genomic segments corresponding to the coding portions of the exons. Introns are excluded from the mature coding sequence because they are removed during normal RNA splicing. For genes with multiple coding exons, the CDS is reconstructed by joining the annotated coding segments in the correct transcript orientation and order. The resulting sequence should preserve the correct reading frame and produce the expected protein sequence. The CDS may begin or end within an exon, so it should not be assumed that an entire exon is coding.
- The distinction between CDS and exon is particularly important in genome annotation. An exon is defined by its presence in the mature transcript, whereas a CDS is defined by its role in protein translation. A single exon may contain both UTR and coding sequence, or it may consist entirely of untranslated sequence. Similarly, a gene may have several exons, but only some of them may contribute to the protein-coding region. Accurate annotation must therefore distinguish transcript structure from protein-coding structure.
- The identification of exon, intron, UTR, and CDS regions involves several complementary approaches. Ab initio gene prediction uses computational models to identify genes based on sequence characteristics, such as start and stop codons, splice-site signals, codon usage, nucleotide composition, and statistical patterns associated with coding regions. Tools such as AUGUSTUS, GeneMark, and other gene predictors use probabilistic or statistical models to infer gene structures. These methods are useful when experimental transcript evidence is limited, although predictions may be uncertain in poorly characterized regions or in organisms with unusual gene structures.
- Homology-based annotation uses similarity to previously characterized genes, transcripts, or proteins from the same or related organisms. BLAST, DIAMOND, and related sequence-search tools can identify homologous sequences that provide evidence for gene boundaries and coding regions. Protein alignments can help locate conserved coding exons, while transcript or cDNA alignments can provide evidence for exon boundaries and splice junctions. However, evolutionary differences, gene duplication, incomplete sequence records, and incorrect annotations in reference databases can affect the accuracy of transferred gene structures.
- Evidence-based annotation uses experimental data to support the identification of transcript structures. RNA sequencing (RNA-seq) provides reads derived from expressed RNA molecules. When these reads are aligned to a reference genome using splice-aware aligners such as STAR or HISAT2, they can reveal exon boundaries, splice junctions, and expressed transcript regions. Transcript assembly tools such as StringTie can reconstruct transcript models from aligned reads, although the resulting models may require further validation. Long-read technologies, including PacBio Iso-Seq and Oxford Nanopore sequencing, can provide evidence spanning multiple exons and are particularly useful for identifying complete transcript isoforms.
- Other experimental evidence can also improve annotation. Full-length cDNA sequencing, expressed sequence tags, cap analysis of gene expression, and 5′ or 3′ end sequencing can help identify transcription start sites and transcript termination regions. Ribosome profiling can provide evidence about translation and help refine coding-region annotations. Proteomics data, including mass spectrometry-derived peptide evidence, can support predicted protein-coding regions, although the absence of detected peptides does not demonstrate that a gene is noncoding or incorrectly annotated.
- Alternative splicing creates additional complexity in exon and intron annotation. A single gene can produce multiple transcript isoforms through exon skipping, alternative 5′ splice sites, alternative 3′ splice sites, mutually exclusive exons, and intron retention. These events can produce transcripts with different exon combinations, coding sequences, and UTR structures. Some isoforms encode proteins with different amino acid sequences, while others may differ primarily in untranslated regions or may be subject to nonsense-mediated mRNA decay. Annotation should therefore represent transcript-specific exon structures rather than assuming that each gene has only one transcript.
- Alternative promoter usage can produce transcripts with different transcription start sites and 5′ UTRs. Similarly, alternative polyadenylation can generate transcripts with different 3′ ends and 3′ UTR lengths. These mechanisms can influence gene expression, RNA stability, localization, and translation. In some cases, alternative transcript structures also affect the coding sequence. A comprehensive annotation should record distinct transcript isoforms and their corresponding exon, CDS, and UTR coordinates whenever supported by evidence.
- Gene annotation pipelines combine prediction algorithms, sequence similarity, and transcript evidence to generate structural annotations. Commonly used tools include BRAKER, MAKER, and other genome annotation systems. BRAKER integrates RNA-seq and protein evidence with gene prediction tools, while MAKER combines evidence from different sources to produce and refine gene models. For prokaryotic genomes, tools such as Prokka and Bakta can identify coding sequences and other genomic features, although prokaryotic genes generally have different exon–intron organization from typical eukaryotic genes. Annotation strategies must be selected according to the organism, genome quality, available evidence, and research objectives.
- The resulting annotations are commonly stored in GFF3 or GTF files. These formats describe genomic features using coordinates, feature types, strand information, and attributes that link genes, transcripts, exons, and coding regions. A simplified GFF3 example is shown below. The coordinates are illustrative and do not represent a complete biological annotation.
##gff-version 3
chr1 annotation gene 1000 2500 . + . ID=gene1
chr1 annotation mRNA 1000 2500 . + . ID=transcript1;Parent=gene1
chr1 annotation exon 1000 1200 . + . ID=exon1;Parent=transcript1
chr1 annotation CDS 1100 1200 . + 0 ID=cds1;Parent=transcript1
chr1 annotation exon 1500 1800 . + . ID=exon2;Parent=transcript1
chr1 annotation CDS 1500 1800 . + 0 ID=cds2;Parent=transcript1
chr1 annotation exon 2200 2500 . + . ID=exon3;Parent=transcript1
chr1 annotation CDS 2200 2300 . + 0 ID=cds3;Parent=transcript1
chr1 annotation five_prime_UTR 1000 1099 . + . Parent=transcript1
chr1 annotation three_prime_UTR 2301 2500 . + . Parent=transcript1
- In GFF3, the gene feature represents the genomic gene interval, while the mRNA feature describes a transcript produced from that gene. Exon features identify the regions retained in the transcript, and CDS features identify the coding portions. UTR features describe the untranslated portions of the transcript. The Parent attribute links a feature to its parent gene or transcript, while ID provides a unique identifier for a feature. The phase column for CDS features indicates how many bases at the beginning of the feature should be skipped to reach the next complete codon, according to the GFF3 convention. Phase values are 0, 1, or 2 and must be interpreted in the direction of translation. In this simplified example, the phase values are illustrative; a real annotation must calculate them consistently across all CDS segments.
- GTF files use a related tabular structure and commonly represent genes, transcripts, exons, and CDS features using attributes such as gene_id and transcript_id. A simplified GTF example is shown below:
chr1 annotation exon 1000 1200 . + . gene_id “gene1”; transcript_id “transcript1”;
chr1 annotation CDS 1100 1200 . + 0 gene_id “gene1”; transcript_id “transcript1”;
chr1 annotation exon 1500 1800 . + . gene_id “gene1”; transcript_id “transcript1”;
chr1 annotation CDS 1500 1800 . + 0 gene_id “gene1”; transcript_id “transcript1”;
chr1 annotation exon 2200 2500 . + . gene_id “gene1”; transcript_id “transcript1”;
chr1 annotation CDS 2200 2300 . + 0 gene_id “gene1”; transcript_id “transcript1”;
- GFF3 and GTF are widely used in genome annotation pipelines, but their conventions are not identical. GFF3 uses explicit feature identifiers and parent–child relationships, whereas GTF commonly relies on gene and transcript identifiers in attributes. Both formats use one-based, closed genomic coordinates, meaning that the start and end positions are included. This differs from some programming languages and interval formats that use zero-based, half-open coordinates. Careful coordinate conversion is essential when comparing annotations or extracting sequences.
- After annotation, exon, intron, UTR, and CDS features should be validated for biological consistency. Important checks include verifying that exons are within their parent gene, transcript coordinates are consistent with strand orientation, introns correspond to gaps between consecutive exons, and CDS segments form a valid coding sequence. The reconstructed CDS should have a biologically plausible start and stop codon where expected, and its length should be consistent with the reading frame. Exceptions may occur in partial genes, incomplete transcripts, organellar genomes, or genes with non-standard translation codes, so validation rules should account for the biological context.
- Splice junctions can be assessed using RNA-seq evidence, while transcript boundaries can be checked using full-length cDNA or long-read data. Protein alignments and conserved domains can help identify incorrectly predicted coding regions. Annotation quality can also be evaluated using tools such as AGAT for GFF/GTF processing and validation, gffread for transcript and coding-sequence extraction, and genome browsers such as IGV and JBrowse for visual inspection. BUSCO can provide information about the completeness of conserved gene sets, although it does not directly validate every exon, intron, or UTR boundary.
- A basic example of extracting transcript sequences and coding sequences from a genome and annotation file using gffread is shown below:
# Extract transcript sequences from a genome and GFF annotation
gffread annotation.gff3 -g reference_genome.fa -w transcripts.fa
# Extract coding sequences
gffread annotation.gff3 -g reference_genome.fa -x cds.fa
# Extract predicted protein sequences
gffread annotation.gff3 -g reference_genome.fa -y proteins.fa
- These commands depend on a compatible annotation file and reference genome. The genome assembly used for extraction must match the assembly against which the annotation coordinates were defined. Otherwise, extracted sequences may be incorrect or incomplete. The resulting transcript, CDS, and protein files can be examined for sequence completeness, unexpected stop codons, reading-frame disruptions, and consistency with available biological evidence.
- Accurate exon, intron, UTR, and CDS annotation has several important applications. In gene expression analysis, transcript and exon annotations help quantify RNA-seq reads and investigate differential exon usage. In alternative splicing studies, transcript-specific exon structures reveal changes in RNA processing between tissues, developmental stages, or experimental conditions. In variant interpretation, variants affecting splice sites, coding regions, or UTR regulatory elements can be investigated for their potential effects on gene function and expression. In comparative genomics, exon–intron organization can help identify conserved gene structures, evolutionary changes, and differences among species. In functional genomics, UTR annotations support the study of post-transcriptional regulation, while CDS annotations provide the sequences needed for protein prediction and functional analysis.
- Several challenges can reduce annotation accuracy. Incomplete or fragmented genome assemblies may prevent correct reconstruction of long genes and transcripts. Repetitive regions can complicate alignment and gene prediction, while lowly expressed genes may lack sufficient RNA-seq evidence. Short-read sequencing may not resolve complex alternative splicing patterns because individual reads often cover only a small part of a transcript. Long-read sequencing can improve transcript reconstruction but may contain platform-specific errors and may not provide uniform coverage. In addition, transcript structures can vary across tissues, developmental stages, environmental conditions, and individuals, meaning that a reference annotation may not capture every transcript expressed in a particular biological context.
- Another important challenge is distinguishing genuine biological complexity from technical artefacts. Incorrectly aligned reads, genomic DNA contamination, incomplete transcript assemblies, and errors in reference sequences can produce false exon boundaries or unsupported transcript models. Annotation transferred from related species may also contain errors if gene structures have changed during evolution. For this reason, predicted features should be supported by multiple independent evidence sources whenever possible, and uncertain regions should be identified rather than assigned unjustifiably precise boundaries.
- Reproducible annotation requires documentation of the reference genome assembly, annotation release, software names and versions, parameters, evidence datasets, and filtering criteria. Researchers should preserve the original annotation files and record any modifications made during analysis. When comparing annotations, they should ensure that genomic coordinates refer to the same assembly and that transcript identifiers and version numbers are handled consistently. These practices make it easier to reproduce gene models, interpret differences between annotation releases, and assess the reliability of downstream analyses.
- Exon, intron, UTR, and coding sequence annotation provides the detailed structural information needed to understand how genes are organized and expressed. Exons define the regions retained in mature transcripts, introns identify intervening sequences removed during splicing, UTRs contribute to post-transcriptional regulation, and CDS regions specify the protein sequence. Together, these features form the foundation of accurate transcript annotation, gene expression analysis, protein prediction, variant interpretation, and comparative genomics. Their reliable identification depends on integrating computational prediction with sequence similarity, transcript evidence, experimental validation, and careful quality control.