Sequence Identity Calculator: Percent Identity and Similarity of Two Sequences

How it works

identity = identical columns ÷ D, where D = alignment length (gaps count), ungapped columns, shorter, mean or longer sequence length; similarity = (identical + conservative) ÷ alignment length; p = mismatches ÷ ungapped columns; Jukes–Cantor d = −¾ ln(1 − 4p/3); Kimura d = −½ ln(1 − 2P − Q) − ¼ ln(1 − 2Q); Poisson d = −ln(1 − p)

Percent identity is the share of alignment columns in which both sequences have the same residue, but the result depends on the denominator, so a figure is only meaningful when you say which one you used (May 2004; Raghava and Barton 2006). Dividing by the full alignment length counts every gap against identity; dividing by ungapped columns ignores gaps, which is close to what BLAST reports for the aligned region; dividing by the shorter, mean or longer sequence describes how much of whole sequences match. The calculator gives all five and lets you choose the headline one. You can paste sequences that are already aligned (with “-” for gaps, from BLAST, Clustal, MUSCLE or a FASTA alignment), or plain sequences, which are aligned here with a global Needleman–Wunsch alignment with affine gap penalties, free end gaps and BLOSUM62 for proteins (match +5, mismatch −4 for DNA), the same defaults as EMBOSS Needle. Very long pairs should be aligned with a dedicated tool first. Protein similarity counts pairs with a positive BLOSUM62 score, as EMBOSS does, or, if you prefer, residues in the same chemical group. For DNA the transition/transversion count and ratio (often about 2 in real genes) check that an alignment is plausible, and the p-distance is corrected for multiple hits with the Jukes–Cantor and Kimura two-parameter models; for proteins a Poisson correction is used. Identity alone does not prove homology: short alignments and low-complexity regions can look similar by chance.

Worked example

The default pair is already aligned: 34 columns, with one gap in sequence 1 at position 23 (33 and 34 residues). 33 columns are identical, 0 mismatched and 1 is a gap, so identity over the alignment is 33 ÷ 34 = 97.1%, over ungapped columns 33 ÷ 33 = 100% and over the shorter sequence 33 ÷ 33 = 100%; over the mean length (33.5) it is 98.5% and over the longer one 97.1%. The gap is internal, so it costs one opening penalty: the BLOSUM62 score of the 33 matches is 160, minus 10 = 150.0. With no mismatches the p-distance and Poisson distance are both 0.

Sources: Raghava & Barton (2006), Quantification of the variation in percentage identity for protein sequence alignments, BMC Bioinformatics · Biostrings pid(): PID1–PID4 definitions (after May 2004) · Sequence Manipulation Suite – Ident and Sim · Biostars: how EMBOSS Needle reports identity and similarity

Frequently asked questions