Global vs Local Sequence Alignment: What Each Job Is For

Global versus local sequence alignment is a job split, not a brand split. Global alignment asks whether two records correspond from beginning to end. Local alignment asks only where a similar stretch exists. NCBI's Global Align help states the contrast in those terms. BLAST is a local search tool. It is not a third kind of sequence, and this page is not a BLAST tutorial.

Two Jobs, Not Two Brands

NCBI's wording is the definition to keep. A global alignment aligns two sequences from beginning to end, each letter used once, and an alignment is produced whether or not the sequences are similar. A local alignment aligns only the portions that share similarity. If there is no similarity, a local method may return nothing.

Needleman–Wunsch is the usual name for the global algorithm. Smith–Waterman is the usual name for the exact local algorithm. Those names pick a scoring procedure. They do not pick a vendor. Semi-global alignment — ignoring end gaps — is a third job and is out of scope here.

When the Whole Sequence Has to Correspond

Use global alignment when you expect the two records to be related along most of their length: two versions of one CDS, two Sanger reads of one amplicon, or two isoforms that should differ by a known internal stretch. NCBI's own example is the dystrophin pair NP_004013 and NP_004014. The first sequence is missing about 100 residues starting at residue 948. A global algorithm returns one alignment that shows the internal difference. A local algorithm returns two alignments, and the change is hard to see.

NCBI's advice matches that example: use global only on sequences that share significant similarity over most of their extents. The job is presentation of whole-record correspondence, not a homology certificate.

When Only a Stretch Has to Match

Use local alignment when you are looking for a conserved domain, a motif, or a short query inside a longer subject. NCBI: only similar portions are aligned. That is the everyday database-search job. BLAST is named Basic Local Alignment Search Tool for that reason. It finds regions of local similarity. It is not a global aligner, and this page will not walk BLAST parameters.

If the question is "where do these two proteins share a domain," local is the job. If the question is "how do these two full CDS records differ along their length," global is the job. Mixing the jobs is how an internal deletion disappears into two local hits.

Do Not Rank Software in Place of the Job

Vendor guides that help you pick Needleman–Wunsch versus Smith–Waterman inside one product are software choosers. This page assigns the job first. After the job is named, a workspace such as ZettaGene is one place that lists alignment among visualization, editing, and construction tools. That is an example of where a lab can run the job. It is not a ranking, and it is not an accuracy claim. Alignment sits inside molecular biology software as one capability, not as the subject of a bake-off.

A Global Picture Can Still Be a Poor Match

NCBI is explicit: a global alignment is produced regardless of whether there is similarity. You still get a picture. That picture is not evidence the sequences are homologs. A local search that returns nothing is the more honest empty result. Do not read a forced end-to-end alignment as relatedness, and do not treat a high-looking identity on a gapped global as a substitute for a local domain hit you actually needed.

Frequently Asked Questions

What is the difference between global and local sequence alignment?

Global asks for end-to-end correspondence. Local asks only for the similar stretch. NCBI produces a global picture even when similarity is poor; a local method may return nothing.

When should I use global alignment instead of local alignment?

When the two records should match along most of their length. NCBI's isoform pair is the case where global makes an internal deletion obvious and local splits it.

Is BLAST a global alignment tool?

No. BLAST is a local search. NCBI's Global Align page is a separate Needleman–Wunsch tool.

Does a global alignment prove two sequences are related?

No. Global still returns a picture when similarity is poor. That is a presentation, not a homology test.