What is Dr. Zubair Khalid's research focus?

Dr. Zubair Khalid specializes in molecular virology, mRNA vaccine development, and computational biology, with a focus on avian pathogens like IBDV and Avian Reovirus.

Where is Dr. Zubair Khalid currently working?

Dr. Zubair Khalid is a Postdoctoral Research Associate at the University of Maryland (UMD), specifically within the Department of Animal and Avian Sciences.

What Is an Expressed Sequence Tag (EST)?

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-09-20

An expressed sequence tag (EST) is a short, single-pass nucleotide sequence, typically 200 to 800 base pairs, read from one end of a randomly picked cDNA clone. Because each cDNA clone is a DNA copy of an mRNA molecule, an EST is a partial snapshot of a gene that was actively transcribed in the tissue the RNA came from.

ESTs matter because they were the first practical way to sample the expressed portion of a genome at scale. Before whole-genome sequencing was routine, EST projects gave biologists a cheap, fast route to discover genes, build the first transcript catalogs for non-model organisms, and generate molecular markers for breeding and conservation. The concept was formalized in the early 1990s, and the National Human Genome Research Institute lists 1991 as the year ESTs were introduced as "fragments of genes" that could be sequenced in bulk to identify expressed regions of the genome [1]. The same logic still explains why EST data remain in public databases and why EST-derived markers are used in plant and animal genetics today.

Why ESTs Changed Gene Discovery

In the late 1980s and early 1990s, finding a new gene usually meant years of work: purify a protein, obtain partial amino acid sequence, design degenerate primers, screen a library, and walk along the chromosome. EST sequencing inverted that workflow. Instead of hunting one gene at a time, a lab could pick hundreds or thousands of cDNA clones at random, sequence a few hundred bases from each, and let the collection of short reads reveal which genes were expressed.

The economic argument was simple. A single-pass read costs a fraction of a fully finished sequence because it is read once, not assembled and verified from multiple overlapping reads. A project that generated thousands of ESTs could produce a working catalog of expressed genes for a species that had no genome assembly, no annotation, and often no funding for a large sequencing center. This is exactly what happened in plants, animals, and fungi. A date palm EST project generated 6,943 high-quality ESTs from a normalized cDNA library of the Sukkari cultivar, assembled them into 6,362 unigenes, and found that 27 percent of the unigenes had no significant similarity to known plant genes, marking them as novel candidates for trait studies [2]. A coralloid root library from the cycad Cycas debaoensis produced 5,011 high-quality ESTs from 5,393 randomly picked clones, which collapsed into 2,984 unique sequences and yielded 22 ESTs encoding putative receptor-like protein kinases plus 94 simple sequence repeats [3].

Those numbers illustrate the central value proposition of ESTs. You do not need a finished genome to learn which genes a tissue expresses, and you do not need to sequence a transcript end to end to recognize it.

The EST Workflow, Step by Step

EST generation follows a fixed sequence of molecular biology steps. Each step introduces a specific kind of bias or error, and understanding the workflow is the fastest way to understand the limits of EST data.

Step 1: mRNA isolation

Total RNA is extracted from a tissue, developmental stage, or stress condition of interest. Polyadenylated mRNA is then purified, usually by oligo(dT) selection that captures the poly(A) tail at the 3' end of eukaryotic transcripts. The choice of tissue determines what the library can reveal. A leaf library will not contain root-specific transcripts, and a library made from unstressed tissue will miss stress-responsive genes unless the plant was challenged first. The Arabidopsis pumila salt-tolerance project is a clean example: the cDNA library was built from leaf tissue that had been exposed to a high-salinity shock, which is why the resulting 16,014 high-quality ESTs were enriched for salt-responsive candidates [4].

Step 2: Reverse transcription to cDNA

Reverse transcriptase copies each mRNA strand into complementary DNA (cDNA). The enzyme needs a primer, and the primer choice determines which end of the transcript is captured. Oligo(dT) priming starts at the poly(A) tail and produces the 3' end of the transcript first. Random hexamer priming starts at many internal positions and gives better coverage of long transcripts and the 5' end. Many libraries combine both. Some protocols add a template-switching step that appends a known sequence to the 5' end, which is how full-length enriched libraries are built. The earthworm Eisenia fetida library used Switching Mechanism at 5' End of RNA Template technology to enrich for full-length clones before sequencing [5].

Step 3: Cloning into a vector

The cDNA is ligated into a plasmid or phage vector and introduced into Escherichia coli by transformation. Each bacterial colony that grows carries one cDNA insert. The result is a library, a living archive of the transcripts present in the original sample. Library quality is judged by titer, the number of colony-forming units per milliliter, and by average insert size. The Cycas debaoensis coralloid root library had a titer of about 1.5 × 10⁶ cfu/mL and an average insert size near 1 kb, with a 97 percent recombination rate [3].

Step 4: Normalization to reduce redundancy

If you sequence a raw cDNA library, the most abundant transcripts dominate. Ribosomal protein genes, storage proteins, and other high-copy messages can account for a large share of clones. Normalization uses duplex-specific nuclease or a related method to equalize transcript representation before cloning. The date palm library was normalized [2], and the Cycas debaoensis library used duplex-specific nuclease normalization combined with template switching [3]. Normalization does not eliminate abundance bias, but it makes rare transcripts easier to find.

Step 5: Random colony picking

Individual colonies are picked at random into multiwell plates. Randomness is the point. If you picked only the colonies that hybridized to a known probe, you would rediscover known genes. Random picking lets the sequencing itself decide what is interesting.

Step 6: Single-pass sequencing

Each clone is sequenced once from one end, usually the 5' end, using a universal primer that sits in the vector just outside the insert. The read is short by design. In the Sanger era, a single pass gave roughly 400 to 800 useful bases. In next-generation EST projects, reads were shorter but far more numerous. The chrysanthemum EST project used 454 pyrosequencing on normalized full-length, 3'-UTR, and 5'-UTR libraries and produced 3,772,677 high-quality reads that assembled into 213,204 contigs [6]. The grass carp project pooled brain, muscle, liver, gonad, and intestinal RNA and generated 738,604 high-quality 454 reads, which assembled into 37,086 all-unigenes [7].

Step 7: Clustering and assembly into contigs

Because many clones represent the same gene, the raw ESTs are clustered by sequence similarity and assembled into contigs, which are contiguous consensus sequences built from overlapping reads. Reads that match no other read stay as singletons. Contigs plus singletons are called unigenes. The Arabidopsis pumila project turned 16,014 ESTs into 8,835 unigenes made of 2,469 contigs and 6,366 singletons [4]. The date palm project produced 6,362 unigenes from 6,943 ESTs, with 494 contigs and 5,868 singletons [2]. A high singleton fraction is normal for EST projects and reflects both rare transcripts and sequencing errors that prevent a read from clustering cleanly.

Step 8: Annotation and mapping

Unigenes are compared against public databases using BLASTX or a similar translated search. Matches to known genes provide functional annotation. Unigenes with no match are flagged as novel or uncharacterized. The Cycas debaoensis dataset annotated 2,333 unigenes in the NCBI non-redundant database and 1,901 in Swiss-Prot, and assigned 1,495 unigenes to 4,082 Gene Ontology terms [3]. ESTs can also be mapped back to a reference genome once one exists, which is how early EST collections became the backbone of gene annotation for many species.

The following flowchart summarizes the main path from tissue to annotated unigene.

flowchart TD
    A[Collect tissue] --> B[Extract total RNA]
    B --> C[Select polyA mRNA]
    C --> D[Reverse transcribe to cDNA]
    D --> E[Normalize library]
    E --> F[Clone into vector]
    F --> G[Pick random colonies]
    G --> H[Single pass sequencing]
    H --> I[Cluster reads]
    I --> J[Build contigs and singletons]
    J --> K[Annotate by similarity]
    K --> L[Map to genome]

What ESTs Are Used For

Gene discovery in non-model organisms

The original use case remains the strongest one. For a species with no genome assembly, an EST project delivers a first-pass gene catalog. A project on the medicinal legume Mucuna pruriens assembled 191 million base pairs of cleaned reads into 67,561 transcripts with a mean length of 626 bp and an N50 of 987 bp, annotated over 80 percent of them, and identified 7,493 simple sequence repeat motifs [8]. The Notopterygium incisum transcriptome project assembled 120,716 unigenes with an N50 of 850 bp, annotated 58 percent of them, and identified 21 unigenes linked to cold tolerance [9].

Genome annotation

ESTs provide experimental evidence that a predicted gene is actually transcribed. When a genome is annotated, transcript evidence from ESTs and RNA-seq is used to define exon boundaries, confirm splice sites, and distinguish real genes from spurious open reading frames. In cassava, annotations had been built mainly from fragmentation-based methods such as ESTs and short-read RNA-seq, and a later full-length cDNA sequencing project found 615 unannotated alternative splicing events and 421 unannotated loci that the fragment-based annotations had missed [10]. That result is a direct demonstration of what EST-level data can and cannot resolve.

Molecular marker development

ESTs are a rich source of simple sequence repeats (SSRs), also called microsatellites, which are short tandem repeats that vary in length between individuals. Because EST-SSRs sit inside transcribed genes, they are often conserved across related species, which makes them transferable. This is the single largest modern use of EST data.

The blackcurrant study identified 14,258 EST-SSR loci from 9,531 coding sequences longer than 1 kb, comprising 6,211 mononucleotide, 4,277 dinucleotide, and 2,469 trinucleotide repeats, then validated 31 polymorphic markers across 107 Ribes accessions [11]. A study of five endemic Taiwanese Hynobius salamander species developed cross-amplifying EST-SSRs with an interspecies transferability rate of 67.67 percent and a polymorphic information content of 0.63, enough to classify 140 individuals into five distinct clusters [12]. Similar work has produced EST-SSR panels for the near-threatened halophytes Artemisia fukudo and Limonium tetragonum [13][14], for endangered Tulipa species in Kazakhstan [15], and for raspberry, where 1,149 red raspberry and 2,358 black raspberry cDNA sequences yielded 121 and 257 SSR loci respectively [16].

ESTs also yield single nucleotide polymorphisms. The Notopterygium incisum dataset produced 85,681 SNPs alongside 13,149 SSRs [9], and the grass carp dataset produced 3,715 SSRs and 2,008 SNPs [7].

Comparative and functional genomics

Because EST-SSR markers are anchored in genes, they can be compared across genera and higher taxonomic groups. The raspberry project was framed around exactly this property, noting that SSRs show relatively high codominant inheritance and transferability within and among taxonomic groups, which makes them useful for comparative mapping and for delineating genomic regions tied to traits of economic importance [16].

ESTs Compared With Full-Length cDNA Sequencing and RNA-seq

ESTs, full-length cDNA sequencing, and RNA-seq all read the transcriptome, but they answer different questions at different costs.

Feature	EST	Full-length cDNA sequencing	RNA-seq
Read length per clone	200 to 800 bp, single pass from one end	Entire insert, typically 1 to 3 kb or more	Short reads, commonly 50 to 300 bp, or long reads on newer platforms
Coverage of a transcript	Partial, one end only	Complete transcript including UTRs and splice junctions	Depends on depth, fragments rather than whole molecules
Typical cost profile	Lowest per clone, high per gene discovered	Highest per transcript, because each clone is fully resolved	Lowest per base, scales with sequencing depth
Main bias	Abundance bias toward high-copy transcripts, plus cloning bias	Abundance bias and cloning bias remain	Abundance bias, plus library prep and fragmentation bias
Error profile	High, because single-pass reads are unverified	Low, because reads are assembled and verified	Low per base, but assembly errors in de novo projects
Best use	Gene discovery, marker mining, annotation evidence	Precise transcript structure, alternative splicing, rare transcripts	Expression quantification, differential expression, variant calling

The practical distinction comes down to resolution. An EST tells you that a transcript exists. A full-length cDNA read tells you what the transcript looks like end to end. RNA-seq tells you how much of it there is and how that amount changes between conditions.

The cassava study makes the resolution gap concrete. Its authors noted that cassava annotations were mainly determined from fragmentation-based sequencing such as ESTs and short-read RNA-seq, then fully sequenced a cDNA library and recovered 8,628 non-redundant full-length transcripts, including rare transcripts that fragment-based methods had missed [10]. The unannotated alternative splicing events they found tended to truncate functional protein domains, and the unannotated loci tended to come from orphan genes, which suggests the missed transcripts were not trivial.

Common Mistakes and Limitations

Treating an EST as a full-length transcript. An EST is a fragment. A 500 bp read from the 5' end of a 3 kb mRNA tells you the transcript exists and gives you its first few hundred bases. It does not give you the coding sequence, the 3' untranslated region, or the exon structure. Any claim about protein sequence based on an EST alone is provisional.

Forgetting the error rate. Single-pass sequencing has no built-in proofreading. Base calls are made once, and errors are not corrected by overlapping reads unless the EST is later assembled into a contig. This is why EST-derived sequences are marked as single-pass in public databases and why consensus contigs are more reliable than individual reads.

Ignoring abundance bias. Even with normalization, EST libraries overrepresent abundant transcripts. A gene expressed at 1 copy per cell may never appear in a library of a few thousand clones, while a gene expressed at 1,000 copies per cell may appear dozens of times. Absence of an EST is not evidence that a gene is not expressed. Normalization reduces this problem but does not eliminate it, and the date palm and cycad libraries both relied on normalization for exactly this reason [2][3].

Assuming EST counts measure expression. EST abundance correlates loosely with transcript abundance, but cloning efficiency, library normalization, and sequencing depth all distort the relationship. ESTs are not a quantitative expression assay. RNA-seq with biological replicates is the correct tool for that question.

Confusing ESTs with genomic sequence. ESTs come from cDNA, so they contain no introns. A BLAST match between an EST and a genomic region will show gaps where introns sit. This is a feature, not a bug, and it is one of the ways ESTs help define exon boundaries during annotation.

Overlooking the vector and quality trimming steps. Raw reads contain vector sequence, poly(A) tails, and low-quality ends. These must be trimmed before clustering. Projects that skip quality filtering inflate their contig counts with chimeric assemblies. The Eisenia fetida project used Phred and Crossmatch to process EST data and retained 1,140 high-quality sequences from its library [5].

Expecting EST data to resolve rare isoforms. Alternative splicing produces multiple transcripts from one gene. An EST from one end of one clone usually cannot distinguish isoforms that share that end. Full-length cDNA sequencing or long-read RNA-seq is required.

Treating EST-SSR transferability as guaranteed. Cross-species amplification rates vary widely. The Hynobius panel achieved 67.67 percent transferability across five species [12], which is useful but far from complete. Every new species pair needs empirical validation.

ESTs in Practice: A Worked Example

Suppose you want to study salt tolerance in a wild relative of a crop plant with no genome assembly. You grow the plants, apply a salt shock, and harvest leaf tissue. You extract total RNA, select poly(A) mRNA, and reverse transcribe with oligo(dT) plus a template-switching oligonucleotide. You normalize the cDNA with duplex-specific nuclease, ligate into a plasmid vector, and transform E. coli. You pick 20,000 colonies into 384-well plates.

You sequence 20,000 clones from the 5' end with a universal vector primer. After trimming vector and low-quality bases, you keep 16,000 high-quality ESTs. You cluster them at 95 percent identity and assemble contigs. You get roughly 2,500 contigs and 6,500 singletons, for about 9,000 unigenes. You run BLASTX against the non-redundant protein database and find that about 8,000 unigenes match known genes while 400 to 500 are uncharacterized. You run Gene Ontology enrichment and find that binding, catalytic, transporter, and stress-response categories are overrepresented, which is consistent with a salt-shock library. You mine the unigenes for SSRs and design 100 primer pairs, then validate 20 that amplify cleanly and are polymorphic across your accessions.

That is the full arc of an EST project, and it mirrors published work closely. The Arabidopsis pumila project followed this path from a salt-shock leaf library through 16,014 ESTs to 8,835 unigenes and functional classification [4]. The date palm project followed the same logic from a normalized library to 6,362 unigenes with stress and fruit development candidates [2].

Quick Review

An EST is a short single-pass sequence, typically 200 to 800 bp, from one end of a randomly picked cDNA clone.
The workflow runs from tissue to mRNA to cDNA to cloning to random picking to single-pass sequencing to clustering and annotation.
ESTs are partial by design, so they identify transcripts but do not define their full structure.
Single-pass reads carry higher error rates than assembled or consensus sequences.
EST libraries are biased toward abundant transcripts even after normalization.
EST-SSRs are the most durable modern use of EST data because they transfer across related species.
Full-length cDNA sequencing and RNA-seq resolve what ESTs cannot: transcript structure, isoform diversity, and expression level.

Frequently Asked Questions

What does expressed sequence tag mean?

An expressed sequence tag is a short sequence read from one end of a cDNA clone, which represents an mRNA that was expressed in the sampled tissue. The word "tag" reflects its role as a marker that identifies a transcript rather than describing it completely.

How long is an EST?

ESTs are typically 200 to 800 base pairs. The exact length depends on the sequencing chemistry used, with Sanger-era reads near the upper end of that range and some next-generation reads shorter.

Are ESTs still used today?

Yes, though the emphasis has shifted. New EST projects are uncommon because RNA-seq and long-read sequencing are cheaper and more informative, but EST data in public databases remain valuable for annotation, comparative genomics, and marker development, and EST-SSR markers are still developed and validated in current studies [12][13][11][15].

What is the difference between an EST and a cDNA?

A cDNA is the DNA copy of an mRNA molecule, and it can be full length. An EST is a short sequence read from one end of a cDNA clone. Every EST comes from a cDNA, but a cDNA is not an EST until it has been sequenced in part.

Why are ESTs biased toward abundant transcripts?

Because library construction samples mRNA molecules in proportion to their abundance, and abundant transcripts are simply more likely to be cloned and picked. Normalization reduces this bias by equalizing representation before cloning, but it does not remove it entirely.

Can ESTs be used to measure gene expression?

Only crudely. EST counts reflect a mix of true transcript abundance, cloning efficiency, normalization, and sequencing depth, so they are not a substitute for RNA-seq with biological replicates when the question is quantitative.