# Ensembl Sequence Export Formats: Which File Type Should You Choose

FASTA gives you raw sequence text, EMBL and GenBank give you sequence plus a human-readable feature table, and GTF, GFF3, and BED give you feature coordinates without the sequence itself. That single distinction settles most format decisions: pick FASTA when a tool needs the letters of DNA, RNA, or protein, pick GTF or GFF3 when a tool needs to know where genes and exons sit on the genome, and pick EMBL or GenBank when you need an annotated, self-describing record. Everything else in this guide is a refinement of that rule.

This article is a practical Ensembl [genome browser tutorial](/blog/guides/genome-browsers-for-researchers-a-guide-to-inspecting-genomic-evidence) focused on sequence export. It walks through the export options the browser, BioMart, and the FTP site offer, explains each format on its own terms with pros and cons, compares them on the criteria that actually decide real choices, and works through concrete scenarios such as primer design, RNA-seq annotation, and [variant effect prediction](/knowledge/bioinformatics/machine-learning-for-variant-effect-prediction-on-protein-stability). By the end you will be able to match any downstream tool to the correct Ensembl export format on the first try.

## How to Use Ensembl: Where Exports Live

Ensembl integrates genomic information for chordate genomes, with deep resources for human, mouse, rat, and zebrafish, and it makes its data available through three main channels: the web browser, BioMart for bulk filtering and export, and programmatic interfaces including a REST API and a Perl API [1]. Non-vertebrate species live on the sister portals under Ensembl Genomes, which cover bacteria, protists, fungi, plants, and invertebrate metazoa through the same browser, FTP, BLAST, and API interfaces [2]. Ensembl Plants, for example, updates four to six times a year and exposes genome sequence, gene models, functional annotation, and variant loci for crop and model species through those shared interfaces [3].

For a first-time user, the export workflow in the browser follows the same pattern regardless of species. The original protocol for using the Ensembl genome server described the core loop that still applies today: search for your gene or protein of interest, open its record, and use the pages dedicated to sequence and comparative data [4]. A later practical tour of the browser emphasized that once you find a gene you can go beyond exporting one sequence and pull integrated data from gene, comparative, and [functional genomics](/blog/guides/functional-genomics) views [5].

In practice you will export sequence from three places:

1. **The Location (region) view.** Navigate to a genomic region, then use the "Export data" link in the left-hand menu. You can choose the sequence format, request the reverse complement, add flanking bases on either side, and mask repeats.
2. **A Gene or Transcript page.** The same "Export data" menu offers genomic sequence, unspliced sequence, cDNA, CDS (the coding sequence, that is, exons joined without introns and without untranslated regions), and the translated peptide sequence. You can toggle UTRs and exons in or out and pick FASTA or an annotated text format.
3. **BioMart.** The data-mining tool supports rapid filtering and bulk retrieval, including exporting the sequences of hundreds or thousands of genes in one query [6]. BioMart is the right choice whenever your input is a list rather than a single locus.

Everything below assumes that mental model: one locus uses the browser menus, a list uses BioMart, and a whole genome or a recurring pipeline step uses FTP or an API.

## Ensembl Sequence Export Formats at a Glance

The table below is the core of this guide. Match your downstream tool to the format column and you have your answer.

| Format | What it contains | Typical consumers | Best for | Main caution |
|--|--|--|--|--|
| FASTA | Raw nucleotide or peptide sequence with a header line per record | BLAST, aligners (BWA, STAR, HISAT2), Primer-BLAST, HMMER, seqkit | Any tool that needs sequence letters only | Contains no annotation, so coordinates and gene structure come from a separate file |
| EMBL | Sequence plus a structured feature table in the INSDC flat-file tradition | INSDC submission pipelines, Artemis, manual record reading | Reading a locus with full annotation, preparing submission-style records | Verbose and line-wrapped, awkward for scripted parsing compared with FASTA plus GFF |
| GenBank | Sequence plus a feature table in the NCBI flat-file style | NCBI tools, Benchling, SnapGene, manual record reading | Working in a NCBI-centric workflow or sharing records with collaborators who expect GenBank | Same parsing awkwardness as EMBL, and interconversion is routine but not lossless for every qualifier |
| GTF | Gene, transcript, exon, and CDS coordinates with attribute fields, no sequence | STAR, HISAT2, featureCounts, StringTie, SnpEff, Cufflinks-era tools | Gene-based RNA-seq and variant annotation pipelines | Must match the genome FASTA assembly and release exactly |
| GFF3 | Feature coordinates in the more general 9-column format, including a directive and sequence option | JKDevice-free browsers, gffread, Apollo, many annotation tools | General feature handling, non-coding features, newer annotation pipelines | Slight syntax differences from GTF trip up scripts written for one or the other |
| BED | Minimal interval coordinates (chromosome, start, end, plus optional fields) | bedtools, [UCSC Genome Browser](/knowledge/bioinformatics/ucsc-genome-browser-beginner-tutorial-and-tips) custom tracks, IGV | Interval arithmetic, intersections with regulatory or experimental data | Uses zero-based, half-open coordinates while Ensembl and GTF use one-based, inclusive ones |
| BioMart TSV/CSV | Whatever columns you select, such as gene IDs, coordinates, cross-references, or sequences | R, Python (pandas), Excel, id mapping steps | Bulk tabular metadata, ID conversion, list-driven sequence export | Column choices are yours, so wrong attribute selections quietly produce useless files |

Two formats in the table deserve a joint note. EMBL and GenBank describe the same kind of content from the two major archives. The European Nucleotide Archive, which combines the Ensembl Trace Archive and the EMBL [Nucleotide Sequence](/knowledge/molecular-biology/nucleotide-sequence) Database, has long prioritized capturing sequence and annotation in a way that survives long-term archiving [7], and EMBL-format flat files remain the submission-facing currency of that archive. GenBank fills the same role at NCBI.

## The Formats in Detail: Pros and Cons

### FASTA

FASTA (Fast-All) is the simplest format in bioinformatics. Each record begins with a header line starting with `>`, followed by the sequence, usually wrapped at 60 or 70 characters. A minimal peptide record looks like this:

```
>ENSP00000269305 pep:BRCA2:ENSP00000269305
MZPI... sequence continues ...
```

**Pros.** Every sequence tool reads it. Files compress well. Ensembl provides genome-wide FASTA on the FTP site in several flavors: unmasked DNA, soft-masked DNA (with repeats in lowercase so repeat-aware aligners can see them), cDNA, CDS, non-coding RNA, and peptide sets. The BioMart interface also emits FASTA for bulk sequence queries, which is how you pull the peptides for an entire gene family in one shot [6].

**Cons.** FASTA carries no coordinates and no gene structure. If you export a genomic FASTA slice and then lose track of the chromosome and strand, the sequence is nearly worthless for genomic work. Header conventions also differ between Ensembl, NCBI, and UniProt files, so scripts that parse headers for one source often break on another. The practical cure is to treat the header as an identifier only and keep metadata in a separate table, which is exactly the discipline large-scale resources follow when they ingest Ensembl sequences into pre-computed similarity databases [8].

A second subtlety is masking. Ensembl's soft-masked DNA files lowercase repeat content while unmasked files uppercase everything. Aligners such as BWA ignore masking, whereas tools built for repeat-sensitive seeding may behave differently on the two versions. Choose one masking convention for a project and record which file you used.

### EMBL flat file format

The EMBL format is a flat-file record that bundles the sequence with a feature table listing genes, exons, CDS locations, translations, cross-references, and free-text qualifiers. Lines are typed by two-letter prefixes such as `ID`, `AC`, `FT`, and `SQ`. This format descends directly from the archiving tradition of the EMBL Nucleotide Sequence Database, where the priority has been to capture sequence and annotation reliably for the long term [7].

**Pros.** One file tells the whole story of a locus. You can read it in a text editor and see the gene structure, the translation, and database cross-references at once. Genome-scale tools such as Artemis render EMBL records graphically. It also matters at the infrastructure layer: GenomeHubs, a containerized system for standing up custom Ensembl browsers, supports export of imported genomes to EMBL format specifically so that assemblies and annotations can be submitted to the International Nucleotide Sequence Database Collaboration [9].

**Cons.** The line-wrapped layout is hostile to quick scripting. Extracting coordinates from an EMBL `FT` line requires more parsing than reading the same fact from GTF. For most pipeline work, an EMBL record is a source you consult rather than a working format you process at scale.

### GenBank flat file format

GenBank is the NCBI counterpart of the EMBL record: sequence plus feature table, with a slightly different syntax (`LOCUS`, `FEATURES`, `ORIGIN`). Biologically it holds the same classes of information, and conversion between the two is routine.

**Pros.** GenBank records open cleanly in NCBI-facing tools and in most desktop sequence editors. If your collaborators think in GenBank, exporting or converting to GenBank removes friction.

**Cons.** The same parsing burden as EMBL. Within Ensembl itself, the browser's annotated text exports are a convenience for reading a locus, while the files designed for pipelines are FASTA plus GTF or GFF3. Treat GenBank and EMBL exports as human-facing records and machine-readable formats as pipeline-facing files.

### GTF (gene transfer format)

GTF is a tab-delimited, nine-column format in which each row describes one feature of a transcript structure. Every row carries the chromosome, the annotation source, the feature type (gene, transcript, exon, CDS, start_codon, stop_codon, five_prime_utr, three_prime_utr), the start and end in one-based inclusive coordinates, a score, a strand, a frame, and an attribute field with key-value pairs such as `gene_id "ENSG00000139618"; transcript_id "ENST00000370046";`. Ensembl distributes a GTF file for every species and release on its FTP site, named on the pattern `Homo_sapiens.GRCh38.<release>.gtf.gz`.

**Pros.** GTF is the lingua franca of gene-level RNA-seq and variant annotation. Spliced aligners consume it to guide transcriptome-aware alignment, quantification tools consume it to count reads over exons, and variant effect predictors consume it to infer transcript structures. The ensembldb R package, which brings Ensembl annotation into Bioconductor workflows for mapping gene-associated data onto coordinates, is built around this Ensembl annotation model and keeps data tied explicitly to Ensembl releases [10]. Variant effect prediction follows the same pattern in practice: published obesity-gene analyses mapped dbSNP variants with SnpEff against the GRCh38.86 reference, pairing Ensembl gene and transcript identifiers with UniProtKB protein accessions to build variant-to-protein relationships [11].

**Cons.** GTF without its matching genome FASTA is dangerous. Coordinates are meaningless against the wrong assembly, and even successive Ensembl releases on the same assembly can shift gene models. The reproducibility answer is to record the release number alongside every file, which is the same release-discipline ensembldb enforces by keeping annotation data separate from software and versioned by release [10].

### GFF3 (Generic Feature Format version 3)

GFF3 generalizes the same nine-column coordinate scheme. It adds a directive line convention (`##gff-version 3`), supports a `###` separator, allows hierarchical parent-child relationships through the `ID` and `Parent` attributes, and can optionally embed the sequence at the end of the file after a `##FASTA` directive. Ensembl distributes GFF3 alongside GTF on the FTP site.

**Pros.** GFF3 represents features GTF cannot express well, including many non-coding and regulatory feature types, and its parent-child model maps naturally onto gene and transcript trees. Newer annotation ecosystems and generic genome browsers accept it broadly.

**Cons.** Some older RNA-seq tools expect GTF strictly. Converting between the two with tools such as gffread or AGAT works, but attribute naming differs enough that blind conversion can drop information. When a tool documentation page says "Ensembl GTF," supply exactly that rather than a converted file, and note that Ensembl Genomes distributes GFF3 through the same FTP structure across all five clades [2].

### BED

![Worked conversion of a 100-base feature starting at position 1,000 from one-based Ensembl coordinates to zero-based half-open BED coordinates.](/article-images/illustrations/ensembl-sequence-export-formats-bed-coordinate-shift.webp)

BED (Browser Extensible Data) is a minimal interval format. The required three fields are chromosome, start, and end, with optional fields for name, score, strand, and more. Its defining quirk is the coordinate arithmetic: BED start positions are zero-based and the end is exclusive, while Ensembl's own displays, GTF, and GFF3 use one-based, inclusive coordinates. Off-by-one errors from mixing these systems are among the most common bugs in interval work.

**Pros.** bedtools, the standard interval arithmetic toolkit, operates natively on BED, so intersecting your exported Ensembl features with ChIP-seq peaks, GWAS loci, or conserved elements becomes one command. The UCSC Genome Browser and IGV both accept BED as custom tracks, which makes it the format of choice when you want to view your Ensembl-derived intervals in a second browser alongside other data, a workflow the UCSC browser documentation describes in detail for visualizing annotations and downloading data for further analysis [12].

**Cons.** BED holds intervals, not gene models. Splice structure is lost unless you emit one BED entry per exon or use BED12 with exon blocks. Converting Ensembl one-based coordinates to BED zero-based coordinates by hand is where most mistakes happen, so prefer a converter over mental arithmetic.

### BioMart tabular outputs (TSV, CSV, HTML) and bulk FASTA

BioMart is Ensembl's query builder. You choose a dataset (for example, Homo sapiens genes), apply filters (a chromosome, a gene list pasted into the filter box, a GO term), select attributes (Ensembl gene ID, external gene name, chromosome, start, end, UniProt ID), and export. Output formats include HTML, TSV, CSV, and FASTA when sequence attributes are selected. Support protocols built around BioMart demonstrate exactly this quick sequence export pattern for disease-gene workflows [6], and the Ensembl Plants toolset describes the same filter-and-retrieve mechanism as its specialist data mining interface [3].

**Pros.** BioMart solves the problem the browser cannot: bulk. Pulling 500 peptide sequences, or a table joining Ensembl transcript IDs to UniProt accessions, takes one query. Cross-references matter more than they first appear, because Ensembl identifiers frequently need translation into other namespaces before they join to protein or chemistry resources, a dependency documented in receptor-focused database reviews that lean on Ensembl for genes and UniProt for proteins [13].

**Cons.** The output is only as good as your attribute selection. Omit the Ensembl gene ID column and you cannot trace rows back to genes. Very large filters can time out in the web interface, at which point the R package biomaRt or a direct FTP download of the whole file becomes the better route.

## Choosing Among Ensembl Sequence Export Formats

Beyond the one-line rule of thumb, five criteria decide real choices. Think of these as the checklist behind the phrase "best Ensembl sequence export formats," because the best format is always relative to a purpose.

**1. Does your tool need sequence, structure, or both?** BLAST and primer design need sequence, so FASTA. STAR and featureCounts need structure, so GTF. SnpEff needs both, a genome FASTA plus a GTF, which is why variant annotation workflows assemble the pair explicitly [11]. If a tool needs both, you almost always need two files from the same release.

**2. What scale are you working at?** One exon, use the browser. One gene list, use BioMart. One genome, use FTP. Whole-genome downloads are routine at Ensembl Plants even for large and polyploid cereal genomes such as wheat and barley, where sequence, gene models, and variant data all ship through the same interfaces [14]. The export options were designed to scale across those tiers, and fighting the wrong tier wastes hours.

**3. Does the coordinate system match?** Ensembl pages, GTF, and GFF3 are one-based and inclusive. BED is zero-based with an exclusive end. VCF is one-based. A 100-base feature starting at position 1,000 appears in Ensembl coordinates as 1,000 to 1,099 and in BED as 999 to 1,100. Make one conversion step, test it on a known feature, and never convert in your head.

**4. Does the release match?** Ensembl re-annotates on a regular release cycle, and gene models between releases on the same assembly can change. The analysis that mapped variants with SnpEff against GRCh38.86 named both the assembly and the release because both matter [11]. Record the assembly, the release, and the file names in your methods, and re-download all paired files together whenever you upgrade anything.

**5. Does the format survive sharing?** If the file will be read by a person, by a reviewer, or by an archive, the annotated flat-file formats win. EMBL format is the export path for INSDC submission in Ensembl-based infrastructures [9]. If the file will be read by a script, tabular or FASTA wins. Matching the audience to the format prevents the most common handoff failures.

## Worked Scenarios

### Scenario 1: Designing primers for one exon of one gene

You want PCR primers flanking a single exon of the human BRCA2 gene (Ensembl gene ID ENSG00000139618). Search the gene in the browser, open the Exons or Genes view, and use "Export data" on the transcript page to pull the cDNA or the specific exon in FASTA. If you want flanking genomic context for primer placement, use the Location view export with a chosen number of flanking bases. Paste the FASTA into NCBI Primer-BLAST. Here the browser export is the correct tier, and FASTA is the correct format, because the primer tool wants sequence only.

The classic browser tutorial flow, find the gene, inspect it, export its sequence, covers this exact loop [4], and the practical guide to the browser explicitly frames single-gene sequence export as the entry-level operation before you move to integrated analyses [5].

### Scenario 2: RNA-seq alignment and quantification

You are quantifying expression with STAR and featureCounts. Download two files from the Ensembl FTP release directory for your species, the primary assembly DNA FASTA and the GTF, and confirm both carry the same release number in their names:

```
curl -O ftp://ftp.ensembl.org/pub/release-113/fasta/homo_sapiens/dna/Homo_sapiens.GRCh38.dna.primary_assembly.fa.gz
curl -O ftp://ftp.ensembl.org/pub/release-113/gtf/homo_sapiens/Homo_sapiens.GRCh38.113.gtf.gz
```

Release numbers change as Ensembl updates, so check the current release directory before scripting. Verify integrity against the checksum files the FTP directory provides. STAR builds its genome index from the FASTA with the GTF supplied via `-sjdbGTFfile`, and featureCounts counts reads using the same GTF. If you work in R and need transcript-to-gene mappings or exon-by-exon coordinate tables, ensembldb builds a versioned annotation resource from Ensembl data that keeps identifier conversion and coordinate mapping consistent within one release [10].

### Scenario 3: Variant effect prediction on a gene panel

You have a VCF of variants in obesity-associated genes and want functional consequences. SnpEff requires a genome FASTA and a GTF for the exact reference your variants were called against. The published obesity analysis did this with GRCh38.86, mapped nonsense and missense effects, and then joined Ensembl gene and transcript IDs to UniProtKB accessions to connect variants to protein-level features such as post-translational modification sites [11]. The workflow has two export takeaways. First, the GTF and FASTA pair is mandatory. Second, the ID mapping step that follows is exactly what BioMart's cross-reference attributes are for, exporting Ensembl-to-UniProt tables that feed tools like Cytoscape downstream.

### Scenario 4: A protein set for BLAST or HMMER searches

You have 300 gene IDs from a literature search and want to build a local BLAST database of their proteins. Use BioMart: filter on your pasted ID list, select the peptide sequence attribute, and export FASTA. This is the documented quick-export use case for BioMart. If instead you want the entire proteome of a species, skip BioMart and download the peptide FASTA from the FTP site, because whole-proteome files are pre-built. Large-scale similarity resources work the same way, ingesting Ensembl protein sequences into pre-computed similarity matrices and clusters that serve many downstream projects [8].

If your comparison work is more bespoke, for example aligning a candidate gene against a genomic scaffold to find exons, the underlying alignment machinery is the same class of tool that genome annotation pipelines use, where heuristic search methods trade a little accuracy for orders-of-magnitude speed against exhaustive alignment [15]. A clean FASTA export is the input half of that bargain.

### Scenario 5: Preparing a new genome for archive submission

You have assembled a non-model species inside an Ensembl-based infrastructure such as GenomeHubs and need to submit the assembly and annotation to the International Nucleotide Sequence Database Collaboration. GenomeHubs supports exactly this handoff through export to flat files, including EMBL format, for INSDC submission [9]. Here the annotated flat file is not optional, it is the archival currency, and the long-term capture priorities of the European Nucleotide Archive are precisely what the format is designed to satisfy [7].

### Scenario 6: Viewing Ensembl intervals in a second browser

You exported a set of regulatory intervals from Ensembl and want to view them in the UCSC Genome Browser next to conservation and ChIP data. Export the intervals as BED from the Ensembl Location view (or convert GTF to BED with a tool), then upload the BED as a custom track at UCSC, where the browser displays tracks around reference assemblies and supports downloading annotations for further analysis [12]. The coordinate conversion in step one is where the one-based to zero-based rule from the checklist earns its keep.

## Step-by-Step Export Walkthrough

![Flowchart branching from the size of the request to the browser export menu, BioMart, the FTP site or the REST API, ending on a release check.](/article-images/illustrations/ensembl-sequence-export-formats-export-route.webp)

For readers who want the browser mechanics spelled out, this compact sequence covers the most common exports.

**Export one genomic region as FASTA.** Search for a coordinate or gene, land on the Location tab, click "Export data" in the left menu, select FASTA, optionally set flanking sequence and reverse complement, and export. The same menu offers EMBL and GenBank-style annotated text for the region if you want the feature table view [5].

**Export one transcript's sequence.** Open the transcript, choose "Export data," and pick among genomic sequence (with or without UTRs), cDNA (exons spliced, UTRs included), CDS (coding bases only), and peptide (the translated product). FASTA is the default and correct output for sequence-only needs.

**Export features as GTF, GFF3, or BED.** From the Location view, the export menu includes feature formats alongside sequence formats. Choose the format your downstream tool names in its documentation.

**Bulk export with BioMart.** Open BioMart, choose the Ensembl Genes dataset and the species, build filters, select attributes, click Results, and export. Choose TSV for scripting and FASTA when sequence attributes are selected,.

**Download whole files over FTP or HTTP.** For recurring pipelines, download directly from the FTP server as shown in Scenario 2. This route also serves the non-vertebrate clades, where Ensembl Genomes exposes FTP, BLAST, and API access across bacteria, protists, fungi, plants, and metazoa through a common interface set [2].

**Programmatic sequence retrieval.** For scripts that need single sequences on demand, the Ensembl REST API returns FASTA directly:

```
curl -H "Content-Type: text/x-fasta" https://rest.ensembl.org/sequence/id/ENSG00000139618 -o brca2_genomic.fa
```

This endpoint pattern is standard for pulling sequence by Ensembl identifier without touching the browser at all, and it complements the fuller Perl and R programmatic interfaces Ensembl has maintained since its early years [1].

## Common Mistakes and Limitations

The recurring failures in Ensembl export work cluster into a short list. Each has a one-line fix.

- **Exporting cDNA when you meant genomic sequence.** A cDNA export has introns removed, so it will not align across splice junctions against genomic DNA and it will give wrong primer results for genomic assays. Check which sequence type you selected before exporting, and remember the toggle for UTRs changes cDNA length.
- **Mixing assemblies.** Human work spans GRCh37 and GRCh38, and coordinates differ between them. A GTF from GRCh38 applied to a GRCh37 FASTA, or variant calls lifted between assemblies without lifting over, produces silent errors. Pin the assembly and release in every file name and methods paragraph, as the variant-annotation literature does when it reports GRCh38.86 [11].
- **Mixing releases.** Gene models move between Ensembl releases even on the same assembly. Pair FASTA and GTF from one release and re-download both together when you upgrade. Versioned annotation resources such as ensembldb exist precisely because release discipline underpins reproducible mapping of gene-associated data [10].
- **Coordinate arithmetic errors at the BED boundary.** One-based inclusive versus zero-based half-open coordinates cause off-by-one shifts. Verify with a feature of known length.
- **Ignoring masking state.** Soft-masked and unmasked DNA files behave differently in repeat-sensitive tools. Note which one you indexed.
- **Header parsing assumptions.** Ensembl FASTA headers are stable and simple, but scripts written against NCBI-style piped headers break on them. Parse IDs defensively or strip headers to the first token.
- **BioMart attribute omissions.** If the exported table lacks the Ensembl gene or transcript ID column, downstream joins become guesswork. Always include the stable identifier columns, and include cross-references such as UniProt IDs when the analysis will touch protein resources [13].
- **Assuming the browser export replaces the FTP file.** Browser exports are per-locus conveniences. Pipelines need the release-level files, and bulk analyses need BioMart or FTP, consistent with how the resource positions its access tiers for browser, API, and BioMart use [1].
- **Treating annotation as ground truth.** Ensembl gene models are computational and curated annotations, not measurements. Alternative transcripts, gene model revisions, and automatic annotation all mean the same locus can legitimately have different-looking models across sources. The browser's strength is integrating external data with automatic annotation so you can inspect rather than assume [4].

One final limitation belongs on the record. Browser menus and export options evolve between releases, so menu names and file paths in any tutorial, including this one, can shift. When a step does not match what you see, consult the release notes and the export dialog itself rather than assuming the workflow is broken.

## Frequently Asked Questions

### What is the difference between FASTA, GFF3, and GTF exports from Ensembl?

FASTA contains raw sequence letters with a header per record. GFF3 and GTF contain feature coordinates and gene structure but no sequence. Use FASTA for tools that read sequence, such as BLAST and aligners, and GFF3 or GTF for tools that read annotation, such as quantification and variant effect tools.

### Which Ensembl export format do I need for RNA-seq quantification?

You need two files, the primary assembly DNA FASTA and the GTF, both from the same Ensembl release. The aligner uses the FASTA, and the GTF guides transcript-aware alignment and exon-level counting. Downloading both from the same release directory is the simplest way to guarantee they match.

### How do I download the whole human genome sequence from Ensembl?

Use the FTP site rather than the browser. Download the primary assembly FASTA file for your chosen release, verify it against the provided checksums, and keep the release number recorded with the file. The FTP directory also holds cDNA, CDS, peptide, and non-coding RNA FASTA files alongside GTF and GFF3 annotation.

### What is the difference between cDNA, CDS, and peptide exports?

cDNA is the spliced transcript including untranslated regions. CDS is the coding sequence only, with UTRs removed. Peptide is the translated protein. Choose cDNA for transcript-level sequence work, CDS when you need codons without UTRs, and peptide for protein searches and domain analysis.

### Can I export sequences for a whole list of genes at once?

Yes, that is what BioMart is for. Paste your gene IDs into a BioMart filter, select the sequence attribute, and export FASTA for hundreds or thousands of sequences in one query. For entire proteomes or transcriptomes, download the pre-built FASTA file from the FTP site instead.

### Do Ensembl and UCSC use the same genome coordinates?

For the same reference assembly, such as GRCh38, the underlying coordinates agree, and data can be moved between the browsers. The practical differences are in file conventions and default annotation sets, so convert formats carefully, watch the BED zero-based coordinate rule, and always record which assembly every file belongs to.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "What is the difference between FASTA, GFF3, and GTF exports from Ensembl?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "FASTA contains raw sequence letters with a header per record. GFF3 and GTF contain feature coordinates and gene structure but no sequence. Use FASTA for tools that read sequence, such as BLAST and aligners, and GFF3 or GTF for tools that read annotation, such as quantification and variant effect tools."
      }
    },
    {
      "@type": "Question",
      "name": "Which Ensembl export format do I need for RNA-seq quantification?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "You need two files, the primary assembly DNA FASTA and the GTF, both from the same Ensembl release. The aligner uses the FASTA, and the GTF guides transcript-aware alignment and exon-level counting. Downloading both from the same release directory is the simplest way to guarantee they match."
      }
    },
    {
      "@type": "Question",
      "name": "How do I download the whole human genome sequence from Ensembl?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Use the FTP site rather than the browser. Download the primary assembly FASTA file for your chosen release, verify it against the provided checksums, and keep the release number recorded with the file. The FTP directory also holds cDNA, CDS, peptide, and non-coding RNA FASTA files alongside GTF and GFF3 annotation."
      }
    },
    {
      "@type": "Question",
      "name": "What is the difference between cDNA, CDS, and peptide exports?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "cDNA is the spliced transcript including untranslated regions. CDS is the coding sequence only, with UTRs removed. Peptide is the translated protein. Choose cDNA for transcript-level sequence work, CDS when you need codons without UTRs, and peptide for protein searches and domain analysis."
      }
    },
    {
      "@type": "Question",
      "name": "Can I export sequences for a whole list of genes at once?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Yes, that is what BioMart is for. Paste your gene IDs into a BioMart filter, select the sequence attribute, and export FASTA for hundreds or thousands of sequences in one query. For entire proteomes or transcriptomes, download the pre-built FASTA file from the FTP site instead."
      }
    },
    {
      "@type": "Question",
      "name": "Do Ensembl and UCSC use the same genome coordinates?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "For the same reference assembly, such as GRCh38, the underlying coordinates agree, and data can be moved between the browsers. The practical differences are in file conventions and default annotation sets, so convert formats carefully, watch the BED zero-based coordinate rule, and always record which assembly every file belongs to."
      }
    }
  ]
}
</script>

## Sources

1. [Ensembl's 10th year.](https://pubmed.ncbi.nlm.nih.gov/19906699/)
2. [Ensembl Genomes: extending Ensembl across the taxonomic space.](https://pubmed.ncbi.nlm.nih.gov/19884133/)
3. [Ensembl Plants: Integrating Tools for Visualizing, Mining, and Analyzing Plant Genomic Data.](https://pubmed.ncbi.nlm.nih.gov/27987162/)
4. [Using the Ensembl genome server to browse genomic sequence data.](https://pubmed.ncbi.nlm.nih.gov/18428779/)
5. [Touring Ensembl: a practical guide to genome browsing.](https://pubmed.ncbi.nlm.nih.gov/20459808/)
6. [Disease and phenotype data at Ensembl.](https://pubmed.ncbi.nlm.nih.gov/21400687/)
7. [Priorities for nucleotide trace, sequence and annotation data capture at the Ensembl Trace Archive and the EMBL Nucleotide Sequence Database.](https://pubmed.ncbi.nlm.nih.gov/18039715/)
8. [SIMAP--a comprehensive database of pre-calculated protein sequence similarities, domains, annotations and clusters.](https://pubmed.ncbi.nlm.nih.gov/19906725/)
9. [GenomeHubs: simple containerized setup of a custom Ensembl database and web server for any species.](https://pubmed.ncbi.nlm.nih.gov/28605774/)
10. [ensembldb: an R package to create and use Ensembl-based annotation resources.](https://pubmed.ncbi.nlm.nih.gov/30689724/)
11. [Systems biology-driven analysis of obesity-associated genes with transcript mapping and post-translational modification annotations.](https://pubmed.ncbi.nlm.nih.gov/42172902/)
12. [Visualizing Genomic Annotations with the UCSC Genome Browser.](https://pubmed.ncbi.nlm.nih.gov/27574198/)
13. [Retrieving GPCR data from public databases.](https://pubmed.ncbi.nlm.nih.gov/27472010/)
14. [Triticeae resources in Ensembl Plants.](https://pubmed.ncbi.nlm.nih.gov/25432969/)
15. [Automated generation of heuristics for biological sequence comparison.](https://pubmed.ncbi.nlm.nih.gov/15713233/)