Plasmid Annotator: Detect Features on a Map

It looks for common cloning elements using short signature sequences: promoters (T7, T3, SP6, lac, tac, araBAD, CMV, EF-1a, SV40), the lac operator and CAP site, Shine-Dalgarno and Kozak sequences, affinity/epitope tags (His6, FLAG, HA, Myc, Strep-II, V5, T7), origins of replication (ColE1/pMB1, f1, SV40), antibiotic-resistance markers (AmpR/bla, KanR, CmR, TetR), universal sequencing primers (M13, T7) and a pUC-style multiple cloning site.

There are two scans. The quick scan flags a feature by matching a short, characteristic signature (roughly 50 bp or less) on both strands, tolerating IUPAC ambiguity codes; because the library is curated and the signatures are short, variants, synonymous codon choices and anything outside the library will not be recognized. Treat it as a fast first pass. The deep scan is the thorough option: it runs BLAST, DIAMOND and Infernal against pLannotate's full databases, so it finds variants and partial features the signature library misses — but because it reports imperfect matches too, check the percent identity column before trusting a low-scoring hit.

Yes. Every signature is searched on the forward strand and as its reverse complement, so features encoded on the antisense strand are reported with forward-strand coordinates and a minus (reverse) strand label.

Your sequence is never written to a database or logged, and no account is needed. The quick scan runs entirely on SeqBench's own server. A deep scan additionally sends the sequence to SeqBench's own pLannotate annotation service over a private network — it is not sent to any third party — and the result is held in that server's memory for up to 24 hours so an identical repeat request is instant. Nothing is persisted to disk. Both plasmid_annotate and plasmid_deep_annotate are also callable from the REST API and MCP server.

The deep scan calls pLannotate, an open-source reimplementation of the common-features auto-annotation you get from desktop tools like SnapGene, cross-referenced against roughly 195,000 SnapGene-annotated plasmids deposited at Addgene. Instead of matching short signatures it aligns your sequence against full feature databases — SnapGene common features, Swiss-Prot, FPbase and Rfam — using BLAST, DIAMOND and Infernal. On pUC19 it recovers features the quick signature scan has no entry for at all, such as the RNAI transcript, the bom site and the rop coding sequence. It takes a few seconds rather than being instant.

Deep scan reports imperfect and truncated matches rather than only exact hits, which is what lets it recognize engineered variants. Identity is the percent match over the aligned region; coverage is how much of the reference feature is present, and a "partial" badge means only part of it is. A hit at 99-100% identity with full coverage is a confident call; something in the 60-70s may be a distant homolog rather than the named feature, so treat it as a lead to check, not a conclusion.

Yes. Deep scan runs a multi-second database search on a shared service, so it is capped at 30,000 bp per sequence — which covers essentially every plasmid, including large expression constructs — and is rate limited per user. Normal interactive use will not hit the limit; if you do, the error tells you how long to wait, and the quick scan stays available and unmetered. The sequence must also be A, C, G, T or N: other IUPAC ambiguity codes are rejected rather than silently dropped, because the search engines discard them and every reported coordinate after such a base would be shifted.