Searching

Start a search

Choose New search. Under Query, enter or paste the sequence: "Residue letters only. The query is read as a sequence, never as a pattern; pattern searches are in Sequence tools." The type, nucleotide or protein, is classified when you submit. Nucleotide and protein queries run in separate pipelines.

Under Who this search is for you can name Patent application this search serves (optional) and, as a sequence searcher, Examiner identifier (optional). Search parameters starts at the defaults: "All parameters are at the defaults set by your administrators." Matrix sets the scoring matrix, and Reset all parameters to the defaults puts every parameter back.

Every search runs against the corpus snapshot active when it is submitted, and stays pinned to that snapshot.

Interactive and deep searches

Under Search depth choose one of two depths.

• Interactive is the default. It aligns a limited set of candidates, so sequences outside that set are not in its result. Nucleotide queries are searched on both strands, and coding nucleotide queries also reach the protein index through six-frame translation. An RNA query is translated as its DNA equivalent (each U read as T); the query itself is kept as you entered it. A query too short for a reading frame of 10 amino acids does not reach the protein index, and its result says so. A deep search does not translate the query.
• Deep: "Deep. Scans every sequence in the databases chosen, or states the fraction it covered; it has no out-of-distribution check, no Patent Significance Scores and no profile search. It needs the deep-search permission, uses your deep-search quota and is confirmed before it is sent."

A deep search is sent only after a confirmation, "Send this deep search?", that states its completion estimate or why there is none. Send deep search sends it; Go back and change returns to the form. Queue class for a deep search offers Default or Rush. Rush gives priority for the next available capacity; it does not promise a completion time.

If the answer to a submission is lost, the page says the outcome is unknown. Submitting the same request again cannot run it twice. To run it again as another search, use Submit as a new search.

Choose databases

Databases to search lists the databases of the active corpus snapshot, each with its license, release and, where the snapshot recorded them, its record counts and attribution. All are chosen by default. A search of fewer than all of them says so on its result, and each hit names the database it came from. A choice that is no longer in the snapshot is removed, never widened to other databases.

The search planner does not choose databases for you.

Exclude the application's own filing

Once an application is named, Exclude this application's own filing becomes available: "Removes only the occurrences in this application's own filing; a sequence also found in another filing stays in the result." The exclusion matches occurrences by application number. An occurrence that names only a publication or patent number of that application is not matched and may be listed. The result says how many hits were set aside and, when some of them are identical to the query, how many.

What the identity lane includes

Before any AI tier, an exact-identity lane looks up records identical to the whole query. For a nucleotide query it also looks up the reverse complement. The result states what the lane covers for each query type:

• nucleotide: "Records identical to the whole query or to its reverse complement are always included; a longer record that contains the query is found only by alignment."
• protein: "Records identical to the whole query are always included; a longer record that contains the query is found only by alignment."

A nucleotide query that holds U is looked up as written and also in DNA letters, each U written as T, the way sequence listings write an RNA. The result says whether both forms were looked up, and an identity hit found through the DNA-letter form says so in the part of the query it matches.

Limits for short queries

A record identical to the whole query is listed by the identity lane, however short the query. A longer record that contains a short query is another matter:

• For a nucleotide query of 22 nt or fewer, the alignment search does not find a longer record that contains the query or its reverse complement, or a record that differs from the query. Such a record is never reached, at either depth.
• For a nucleotide query of 23 nt, the alignment search does not find the query inside a longer record.
• For a short protein query, whether it is found inside a longer record is not yet measured.

A search whose query meets one of these limits records the limit, and its result states it word for word. An empty result of such a search does not show that no record contains the query.

My searches

My searches lists your searches, "Searches, newest first", with the query, its depth and its status. A sequence searcher also sees the searches they entered for examiners. Tick Include on the searches you want, then export them together as a diligence record.

A search can be replayed against the snapshot it was pinned to; the replay is a new execution, and the original result is unchanged.