Peptide Sequence Identity vs Similarity: Alignment and Coverage
Learn to distinguish exact sequence matches, scored similarity, alignment coverage and homology, using worked peptide examples.
Sequence identity counts exact residue matches in an alignment; sequence similarity can also recognize substitutions according to a defined scoring system. Both depend on the alignment and its denominator. Coverage tells you how much of each sequence was compared. Homology is a claim of shared evolutionary ancestry, not a percentage. None of these measures alone proves that two peptide materials have identical chemistry or function.
Start with the alignment, not the headline percentage
A sequence comparison places residues into corresponding columns. Only after that arrangement is specified can a reader count identical characters, inspect substitutions or decide which positions entered the calculation. Two percentages that look comparable may describe different residue ranges or different treatment of gaps.
The NCBI BLAST glossary distinguishes identity from positive substitutions and describes similarity in relation to those measures. In a protein-style comparison, an identical pair is the same residue symbol in a column. A conservative substitution is different and may receive a favourable score under the chosen scheme. [1]
For short research peptides, precision is especially useful because each position represents a substantial fraction of the sequence. A single mismatch changes identity by 12.5 percentage points in an eight-column ungapped alignment, compared with 1.25 points in an eighty-column alignment. Those arithmetic facts do not establish which substitution matters more biologically.
If the residue alphabet itself is unfamiliar, begin with how to read peptide sequences. A comparison should preserve the actual symbols before attaching a percentage to them.
A worked example: identity is an exact-match count
Hypothetical example: compare the eight-residue sequences AGKLSFWT and AGKISFYT without gaps, using all eight positions. These sequences were invented for arithmetic and were not selected as real biological homologs.
| Position | Sequence A | Sequence B | Identical? |
|---|---|---|---|
| 1 | A | A | Yes |
| 2 | G | G | Yes |
| 3 | K | K | Yes |
| 4 | L | I | No |
| 5 | S | S | Yes |
| 6 | F | F | Yes |
| 7 | W | Y | No |
| 8 | T | T | Yes |
There are six identical columns out of eight. With this stated denominator, identity is 6 ÷ 8 × 100 = 75%. Leucine and isoleucine are different residue symbols; tryptophan and tyrosine are different too. A structural resemblance between a pair does not turn it into an exact match.
The percentage says nothing about where the differences occur. Two mismatches near a critical recognition feature and two mismatches elsewhere produce the same count. Position-specific evidence is needed to decide how a substitution affects an interaction.
Similarity requires a scoring rule
A similarity measure can recognize relationships among non-identical residues. A simple teaching rule might count L/I and W/Y as acceptable substitutions in the invented alignment above. Under that deliberately chosen rule, all eight columns would be accepted, giving 100% accepted positions while exact identity remains 75%. This is a manual illustration, not a reported BLAST result or a universal similarity definition.
Real sequence-analysis tools use explicit scoring systems. Henikoff and Henikoff’s original BLOSUM work derived substitution matrices from conserved blocks of aligned protein families. Such matrices provide different scores for different residue pairings rather than treating every mismatch as equivalent. [2]
A favourable substitution score is not proof that the two residues are interchangeable at every site. A scoring matrix summarizes a model derived from sequence comparisons; it does not measure the effect of this substitution in this particular peptide assay. Local steric contacts, charge placement and backbone constraints can make a seemingly conservative change consequential.
When a paper reports “similarity,” look for its definition. Does it mean a percentage of positive-scoring columns, a total alignment score or a broader qualitative comparison? Without a rule, the number is incomplete. Without the alignment, even a clearly named rule leaves the compared positions uncertain.
Local and global alignment answer different questions
Global alignment compares sequences across their full lengths, while local alignment focuses on matching regions. EMBL-EBI’s pairwise-alignment documentation explains this distinction and the use of substitution scores and gap penalties. The chosen method optimizes an explicit scoring problem; it does not independently decide the biological meaning of the comparison. [3]
Hypothetical example: take sequence A, AGKLSFWTNVDR, containing 12 residues, and sequence B, AGKLSFGGTTAACC, containing 14. Deliberately inspect only the shared first six residues, AGKLSF. This selected six-column region has 100% identity. It covers 6 of 12 residues in A, or 50%, and 6 of 14 in B, or approximately 42.9%.
That statement is accurate only with its scope attached. Rewriting it as “the sequences are identical” would remove the unexamined residues. It is also not a claim that an alignment algorithm would return exactly this region as its best result; no search was run for this example.
Conversely, a full-length comparison can have 100% coverage while including mismatches. Coverage asks how much sequence participates. Identity asks how many compared columns match exactly. Neither replaces the other.
Gaps make the denominator important
Hypothetical example: place AGKLSF opposite AGK–SF, where the dash represents a gap rather than a residue. There are six alignment columns, five of which contain identical residue pairs. If all six columns enter the denominator, identity is 5/6, approximately 83.3%.
If someone instead reports five matches divided by five residue–residue columns, the result is 100%, but the gap has been excluded from that denominator. These calculations answer differently defined questions. The reader needs the convention, rather than an unexplained percentage that hides the insertion/deletion distinction.
Gaps also affect which alignment is selected. Opening a new gap and extending one can carry different penalties. Adjusting these settings can change the preferred arrangement, especially for very short or repetitive sequences. A useful report retains the alignment and settings so another reader can understand the comparison.
| Metric | Question answered | Information to report |
|---|---|---|
| Identity | How many compared positions match exactly? | Match count, alignment length and gap convention. |
| Similarity | How are additional substitutions evaluated? | Scoring matrix or other explicit rule. |
| Coverage | How much of a sequence entered the comparison? | Query and subject lengths and aligned ranges. |
| Statistical significance | How surprising is the score under the search model? | Statistic, search space, tool and relevant settings. |
Why a short match is not an ancestry certificate
Short strings can recur in unrelated sequences. A repeated motif or a compositionally simple segment can provide a visually persuasive match without establishing an evolutionary relationship between the full molecules. The number of candidate sequences searched also matters when considering how often a match might appear by chance.
RCSB PDB’s sequence-search documentation explicitly warns that short queries can match many sequences and lead to false evolutionary conclusions. Its search output combines identity with aligned regions and other information rather than treating identity alone as a complete interpretation. [4]
For a designed research peptide, a sequence resemblance may be a deliberate construction rather than evidence of common biological ancestry. A copied fragment, a shuffled control and an independently designed short sequence have different histories even if a small aligned region matches. The inference should fit the provenance and the question.
Homology and E-values are not percentages of the same thing
Homology refers to shared evolutionary ancestry. The NCBI glossary treats it as an evolutionary relationship, distinct from measured similarity. Avoid expressions such as “75% homologous” when the observation is actually 75% identity across a specified alignment. The percentage belongs to the count, not to the ancestry claim. [1]
An E-value addresses a different issue again. NLM’s BLAST teaching resource defines it in terms of the expected number of chance alignments reaching a given score or better under the search model. It is not the percentage identity and not the probability that the sequences have the same function. [5]
There is no single identity percentage that proves common ancestry, identical function or experimental interchangeability for every peptide length. A short perfect match and a long well-supported alignment are different evidence. Interpretation must consider aligned length, composition, statistical support and biological context rather than importing a threshold from an unrelated application.
Sequence identity is not complete chemical identity
Ordinary residue strings can omit terminal caps, stereochemistry and crosslinks. Two records can match letter for letter while describing different covalent molecules. A sequence-alignment program can accurately report a perfect match for the strings it received while remaining silent about those omitted features.
Hypothetical example: one record lists an uncapped AGKS chain, and another specifies Ac–AGKS–NH2. If a comparison retains only the four residue letters, those inputs match exactly. The N-terminal acetyl group and C-terminal amide still distinguish the complete structures. The alignment result cannot restore information removed before the comparison.
Similarly, identical cysteine positions do not specify the disulfide pairing. Read linear versus cyclic peptides for connectivity, and the peptide-bond guide for the backbone connection. Analytical identity is another question: a sequence comparison is not a physical measurement of a submitted sample. See what mass spectrometry establishes about peptide identity.
Frequently asked questions
Can similarity be higher than identity?
Yes, when a defined similarity rule accepts some non-identical substitutions in addition to exact matches. Report that rule and the common denominator; the word “similar” alone is insufficient.
Does 100% identity mean the entire sequences are identical?
Only if the comparison actually covers the relevant full sequences under a stated convention. A short local region can be perfectly identical while leaving most residues outside the alignment.
Is 100% query coverage the same as 100% subject coverage?
No. Query and subject can have different lengths. An entire short query can align to only a small part of a longer subject, so both ranges and lengths matter.
Does a conservative substitution guarantee retained activity?
No. It is a sequence-comparison description, not an assay result. The significance of a substitution depends on its position and the molecular interaction being studied.
Can one fixed identity threshold prove that peptides are equivalent?
No universal threshold does that. Equivalence is broader than a sequence count and requires a clearly defined research purpose and appropriate evidence.
What should a useful comparison record include?
Retain both full annotated structures, aligned residue ranges, exact-match counts, gap treatment, scoring method and coverage. For database searches, also retain the search context and statistics. The peptide science hub connects these sequence concepts to chemistry and evidence interpretation.
Cover: AI-generated conceptual editorial artwork, not an exact molecular structure, measured result or a photograph of HealthyWelness staff or facilities.