Sequences

Page version 1.0.0 · For Sequences 0.95.0

How to read a result

One hit of a real search, drawn letter by letter, with its four measures worked out and its place against a date.

The GLP-1 sample result is a real search, saved and shown read-only on the hosted service. A search compares a query, the sequence you search with, against stored sequences, and each stored sequence that matches is a hit. This page walks through one of the sample's hits: how Sequences lined it up with the query, how it scored it, and how a date places it.

A protein is a chain of amino acids, and its sequence writes each amino acid as one letter, such as H for histidine. Each amino acid in the chain is a residue. The sample's query is GLP-1(7-37), a form of the hormone GLP-1 that is 31 amino acids long. Its letters come from the public protein database UniProt: UniProt P01275, residues 98 to 128.

HAEGTFTSDVSSYLEGQAAKEFIAWLVKGRG

It searched Patent proteins, the protein sequences published in patents, with the method Automatic, which ran Smith-Waterman: a comparison of the query with every stored sequence, letter by letter. The search ran on November 12, 2026 and found 5,000 sequences, the most a Smith-Waterman search keeps for one query, in 3,061 patent publications. Choosing a search method compares the methods.

One hit, drawn

An alignment sets the query and a hit side by side so that the letters that match line up. Below is the best alignment of WO 2026013265-A1, published on January 15, 2026: positions 1 to 30 of the query against positions 1 to 30 of the patent's sequence.

  • | Identical: the same amino acid
  • + Similar: a different amino acid that the scoring table scores above zero
  • ≠ Different: a different amino acid that scores zero or less
  • ? X in the hit: an amino acid the patent's sequence does not name
  • · Not aligned: a letter of the query outside the alignment
Table view
Each column of the alignment of the query with the best hit of WO 2026013265-A1
GLP-1 positionQuery positionQuery letterHit letterHit positionKindScore
71HH1Identical8
82AX2X in the hit0
93EE3Identical5
104GG4Identical6
115TT5Identical5
126FF6Identical6
137TT7Identical5
148SS8Identical4
159DD9Identical6
1610VV10Identical4
1711SS11Identical4
1812SS12Identical4
1913YY13Identical7
2014LL14Identical4
2115EE15Identical5
2216GE16Different-2
2317QQ17Identical5
2418AA18Identical4
2519AA19Identical4
2620KR20Similar2
2721EE21Identical5
2822FF22Identical6
2923II23Identical4
3024AA24Identical4
3125WW25Identical11
3226LL26Identical4
3327VV27Identical4
3428KR28Similar2
3529GG29Identical6
3630RR30Identical5
3731GNoneNoneNot alignedNot scored
The best alignment of WO 2026013265-A1 with the query, in blocks of ten columns. Each column shows, from the top, the position in GLP-1 numbering, the query's position and letter, a mark, and the hit's letter and position. The table view lists every column with its score.

Each column pairs one letter of the query with one letter of the hit, and its mark says how the two compare:

  • Identical, marked |: 26 columns hold the same amino acid.
  • Similar, marked +: at query positions 20 and 28, the hit has R where the query has K. These are different amino acids, but the scoring table scores the pair above zero, since related proteins often swap one for the other.
  • Different, marked ≠: at query position 16, the hit has E where the query has G, a pair that scores below zero.
  • X in the hit, marked ?: at query position 2, the hit has X where the query has A. In a patent's sequence listing, the part of the application that lists its sequences, X stands for an amino acid the listing does not name with one of the standard letters, such as a modified one.
  • Not aligned, marked ·: query position 31, the last G, is outside the alignment, which stops at position 30.

Two ways to count positions

A result counts positions from the start of each sequence: the query's from 1 to 31, and the hit's from 1 along its own sequence. The name GLP-1(7-37) says the query is amino acids 7 to 37 of GLP-1 as papers and patents number them, so query position 1 is GLP-1 position 7. The figure shows both: the X at query position 2 is at GLP-1 position 8, the different pair at 16 is at 22, the similar pairs at 20 and 28 are at 26 and 34, and the letter left out is at 37.

How the columns score

Each column gets a score from a scoring table called BLOSUM62, which gives each pair of amino acids a score by how often the pair turns up in related proteins. An identical pair scores that amino acid's own value, such as 8 for H against H.

ColumnsPairEach scoresTogether
26 identicalAn amino acid against itselfIts own value, such as 8 for H135
2 similarK against R24
1 differentG against E-2-2
1 XA against X00
All 30 columnsThe whole alignmentAdded up137

The query against itself, all 31 letters identical, scores 161. The G left out at position 31 adds nothing to the alignment's score, but it counts in that perfect score. A gap, a run of letters on one side with nothing across from them, costs 11 plus 1 for each letter in it. This alignment has none.

The four measures

Sequences shows four measures for each hit. They answer different questions, so on this hit each gives a different number.

MeasureWhat it asksOn this hit
Query MatchHow close is the alignment's score to the query's score against itself?137 of 161: 85.1%
IdentityOf the alignment's columns, gaps included, how many hold the same amino acid on both sides?26 of 30: 86.7%
Query identityOf all the query's letters, how many have the same amino acid across from them?26 of 31: 83.9%
Query coverageHow much of the query does the alignment span?30 of 31: 96.8%

Query Match weighs how alike two amino acids are, so the similar pairs raise it and the different pair lowers it. The other three count only identical letters and lengths. Query identity and Query coverage divide by the whole query, so the letter left out at position 31 lowers both, while Identity counts only the columns the alignment holds.

A row of the list shows the best Query Match and the best Identity among all of a publication's hits, and the two can come from different hits. For WO 2026013265-A1, both come from the hit drawn here.

In your own results, the Alignments tab also gives each hit an E-value: about how many hits this good a search of this size would find by chance alone, so the smaller it is, the less likely the match is chance.

Why it is listed, and where it falls in time

WO 2026013265-A1 is in the list because its sequence matches the query. Its best alignment reads "26 of 30 identical, no gaps." Of the letters that differ, 2 are similar, 1 is different and 1 is an X.

Where a publication falls in time depends on a date you choose, such as the filing date of an application you are checking. A publication is published on its publication date and filed on its application's filing date, and a priority claim names an earlier application whose date it claims. Set Published before to your date, and each publication falls in one of three places:

  • Published before your date. It stays in the list.
  • Published after your date, but filed or claiming priority before it. It stays in the list, marked "Filed before". This takes a US publication, or a WO publication, one filed under the international Patent Cooperation Treaty, whose record gives a filing or priority date before yours.
  • Published on or after your date, and not filed or claiming priority before it. It folds under the list when its sequence is known from before your date, such as from an earlier publication, and Show them brings it back, marked "After date". Otherwise it leaves the list.

Take January 1, 2026 as the date, with Group set to Every publication. WO 2026013265-A1 was published on January 15, 2026, after it, but its record gives a filing or priority date before it, so its row is marked Filed before. The sample shows only the publication date and the mark. In your own results, the publication's pane says "Filed before your date" and shows the filing and priority dates where the record gives them.

With this date, 2,981 of the 3,061 publications stay in the list, and 77 later publications of the same sequences fold under it, in one line: "77 later publications of these sequences are folded. 12 of them are US or WO publications with no filing date on record." The other 3 leave the list, because every matching sequence they hold was first known on or after the date, with no US or WO publication filed or claiming priority before it. Dates in the user guide says how each sequence is dated.

A second hit, for contrast

JP 2002507188-A, published on March 5, 2002, holds all 31 letters of the query unchanged, so all four of its measures read 100.0%. But it is not the first publication of those letters. 1,647 other publications in the result hold all 31 letters of the query unchanged. The earliest is WO 9001540-A, published on February 22, 1990, 12 years before it.

Measure or dateJP 2002507188-AWO 9001540-A
Published2002-03-051990-02-22
AlignmentQuery 1 to 31, hit 1 to 31Query 1 to 31, hit 75 to 105
Query Match100.0%100.0%
Identity100.0%100.0%
Query identity100.0%100.0%
Query coverage100.0%100.0%
Row in the list of every publication351

The measures say how well a sequence matches. Only the dates say which publication held it first. With Group set to Every publication, the list's first order, highest Query Match first, puts rows that tie earliest first, so WO 9001540-A is row 1. When a row holds your query unchanged, look for the earliest row that holds it too.

What Sequences does not decide

Sequences finds the stored sequences that match your query, measures how well each one matches, and dates each publication from its record. It does not decide what a match means for a claim. Whether a publication matters, and from which date, is for you to judge from the publication itself: its claims, its description and its dates. A record may give no filing or priority date, and a sequence listing may write the same amino acid in more than one way, as the X above shows. These results do not establish patentability.

More