BLAST and databases

Why it matters

You have a sequence. What is it, and what is it similar to? BLAST (Basic Local Alignment Search Tool) answers that in seconds by comparing your sequence against public databases holding hundreds of millions of sequences. Technicians use it to confirm a PCR product or Sanger read is the gene they expected, to identify a microbe, and to check that a primer will not bind somewhere unintended. It is free, runs in a browser and needs no account.

Sequence databases

The main public databases are run by an international collaboration: NCBI (US National Center for Biotechnology Information), ENA (European Nucleotide Archive) and DDBJ (DNA Data Bank of Japan), which exchange data daily. At NCBI you will meet:

• Nucleotide and GenBank: submitted DNA and RNA sequences.
• RefSeq: curated reference sequences. Accessions start with a prefix such as NM_ (mRNA), NP_ (protein) or NC_ (chromosome or complete genome).
• Protein: protein sequences.
• Gene: summaries that link a gene's sequences, names and literature.

Each record has an accession number with a version, such as NM_000518.5. Citing the accession and version makes your work reproducible.

The BLAST programs

Use blastn to identify a DNA sequence, blastp for a protein, and blastx when you have DNA and want to know which protein it might encode.

Reading the results

BLAST lists hits (database sequences) with alignments. The columns that matter:

• Query cover: how much of your query is included in the alignment. A 98 percent identity over 20 percent of the query is not a convincing match for the whole sequence.
• Percent identity: identical positions within the aligned region.
• E-value: how many hits this good you would expect by chance in a database this size. Smaller is better: 1e-50 is overwhelmingly significant; values around 1 or higher are noise. It depends on database size and query length.
• Max score (bit score): alignment quality on a normalised scale; higher is better and comparable across searches.
• Accession and description: what the hit is.

Worked example. You BLAST a 600 bp PCR product. The top hit: "Homo sapiens hemoglobin subunit beta (HBB), mRNA", query cover 100 percent, identity 99.8 percent, E-value 0.0. One mismatch over 600 bases. Your product is HBB, and the mismatch is either a sequencing error (check the chromatogram at that position) or a real variant worth following up.

A different case: top hit with query cover 35 percent, identity 92 percent, E-value 2e-20. Part of your sequence matches, the rest does not, which might mean a chimeric product, vector sequence, or a region that is not in the database.

The graphic and alignment views

The graphic summary shows each hit as a bar under your query, coloured by score, so you can see at a glance whether hits cover the whole query. The alignment view shows your query (Query) above the database sequence (Sbjct) with | for identical bases and gaps as -, and gives the coordinates on both, 1-based.

Good habits

• Paste sequence in FASTA format, and trim low-quality ends of Sanger reads first.
• Choose the database deliberately (for example RefSeq for curated references, or the default nucleotide collection to cast a wide net).
• Record the program, database, date and top accession(s) in your notebook: databases grow daily, so a search repeated later can give different results.
• Never upload patient-identifiable data or confidential sequences to a public server; clinical labs use approved, often local, tools.

The NCBI BLAST service and its documentation are at https://blast.ncbi.nlm.nih.gov/Blast.cgi (Checked on 2026-10-23.)

Common mistakes

• Trusting percent identity while ignoring query cover.
• Treating an E-value of 0.5 as a match.
• Using blastn on a protein sequence.
• Forgetting that the top hit's description can be wrong or vague if it is a poorly annotated submission.
• Not recording the database and date.

Quick recap

• NCBI, ENA and DDBJ share sequence data; RefSeq holds curated references; cite accession.version.
• Pick the program by query and database type: blastn, blastp, blastx, tblastn.
• Read query cover, percent identity and E-value together.
• Record the program, database, date and hits; keep confidential data off public servers.

Try it

Open the Do tab to fetch a real reference sequence from NCBI and BLAST it.