Pairwise Sequence Alignment Tool

About the Pairwise Sequence Alignment Tool

Paste two DNA, RNA or protein sequences and get their optimal alignment: global (Needleman–Wunsch), end to end, for sequences related over their whole length, or local (Smith–Waterman), the best-matching region, for a domain inside a protein or a gene in a longer sequence. Gaps are scored the affine way, an opening penalty and a smaller one for each extra residue, and the alignment is found in linear memory (the Myers–Miller method) in a background worker, so sequences up to 20,000 residues (and up to 100 million positions to compare) fit in the browser’s memory, on a phone too.

Proteins are scored with BLOSUM62 or another BLOSUM or PAM matrix, DNA with NUC.4.4, which knows the IUPAC codes, or with your own match and mismatch scores. You get the score, the length, the identity, the similarity and the gaps, counted as EMBOSS needle and water count them, and a colour-coded alignment with residue numbers. With a Pro pass, copy the alignment or download it as text or as aligned FASTA, or unlock this one result. Without a pass the free preview shows the score, length, identity, similarity and gaps in full, with the first rows of the alignment (up to 10).

Common uses

• Comparing a gene or a protein between two species
• Checking a sequencing read or a clone against its reference
• Finding a conserved domain inside two longer proteins with a local alignment
• Working out percent identity and similarity for a report
• Learning how dynamic-programming alignment works, with the textbook examples

How the alignment is scored

• Each aligned pair scores its value in the matrix: in BLOSUM62, a tryptophan with a tryptophan scores 11, an alanine with a glutamate −1.
• A gap of n residues costs open + (n − 1) × extend, with 10 and 0.5 as in EMBOSS.
• Global alignments include every residue of both sequences; by default the gap that opens the alignment and the one that closes it cost nothing, which suits a short sequence aligned into a longer one.
• Local alignments keep only the highest-scoring region; an alignment of unrelated sequences can be empty.
• The score is the total; identity counts identical pairs, similarity pairs with a positive score, and gaps the gap positions, each over the length of the alignment including gaps. The markup line shows | for identical, : for similar and . for other pairs.

Choosing a matrix

BLOSUM matrices were built from blocks of aligned protein segments: BLOSUM62 from blocks clustered at 62% identity, the usual choice. A higher number (BLOSUM80, BLOSUM90) suits closely related sequences, a lower one (BLOSUM45) distant ones. PAM matrices are numbered the other way: PAM30 for close relatives, PAM250 for distant ones. For DNA, NUC.4.4 scores a match +5 and a mismatch −4 and gives partial scores to IUPAC codes such as R or N; match and mismatch scores of your own are the alternative.

How it is computed

The alignment uses Gotoh’s three-state recurrences for affine gaps and Myers and Miller’s divide-and-conquer, which needs memory in proportion to the length of the sequences rather than to their product, so 10,000 × 10,000 residues fit in a few hundred kilobytes. A global alignment with unpenalised end gaps first finds where the best alignment ends and starts, then aligns that stretch; a local one does the same around its best-scoring region. Its unit tests check it against a full-matrix aligner on hundreds of random pairs and reproduce the score and statistics of EMBOSS needle’s documented example.

Sources

• Needleman S.B. and Wunsch C.D., Journal of Molecular Biology 48:443 (global alignment); Smith T.F. and Waterman M.S., Journal of Molecular Biology 147:195 (local alignment); Gotoh O., Journal of Molecular Biology 162:705 (affine gaps); Myers E.W. and Miller W., CABIOS 4:11 (linear space).
• Henikoff S. and Henikoff J.G., PNAS 89:10915 (BLOSUM); the matrices as published by NCBI.
• EMBOSS needle: the gap scoring, the statistics and the example.

Limitations

• Two sequences at a time: for many, use a multiple sequence alignment program.
• Up to 20,000 residues each and 100 million positions to compare (10,000 × 10,000); long alignments take seconds on a computer and longer on a phone.
• When several alignments have the same best score, one of them is shown; the score and the statistics may differ slightly from another program’s choice.
• No statistical significance (E-value) is given: a high score of short or low-complexity sequences can arise by chance.
• Selenocysteine (U), pyrrolysine (O) and J in proteins are scored as X.

Privacy

On this device: nothing is sent

Everything happens in your browser. What you enter or open here is not uploaded or stored by MySmartCoPilot.