Analyse a protein sequence

What this involves

Most sequence analyses fail before computation begins because raw inputs retain alignment characters, stop codons, unrecognised ambiguity codes, or expression tags. Rigorous sequence characterisation begins by stripping FASTA headers, verifying residue alphabets, and separating actual sequence modifications from computational artefacts. Running property calculations on uncleaned sequences introduces systematic errors into molecular weight, isoelectric point, and hydropathy values.

Once cleaned, deterministic physicochemical profiles provide baseline estimates for laboratory handling, including buffer selection, concentration measurement via A280 extinction coefficients, and flags for potential insolubility or disorder. When nucleotide sequences are supplied, reading frame verification and six-frame translation are required before computing any amino acid metrics.

For functional and evolutionary context, the workflow queries curated databases and runs local sequence alignment. By combining search tools with domain annotations and protein language model embeddings, the agent resolves homologues while guarding against false transfers of function in the sequence twilight zone.

How it goes

1. 1Sequence sanitisation and validationInspect and clean raw strings using SeqKit or regex utilities to remove headers, strip gaps and stop codons, uppercase soft-masked residues, and evaluate ambiguity codes like B, Z, or X before downstream processing.
2. 2Deterministic property calculationCompute average-isotopic molecular weight, theoretical pI, GRAVY hydropathy, aromaticity, net charge at pH 7, and extinction coefficients using the Sequence Analytics module.
3. 3Database annotation retrievalQuery UniProtKB live via search_proteins to retrieve verified accessions, gene symbols, organism metadata, and functional snippets for known proteins.
4. 4Homology search and alignmentExecute sequence similarity searches against databases such as UniRef90, UniRef50, or Pfam using BLAST+, DIAMOND, or MMseqs2, reporting bit scores, aligned column counts, and query coverage.
5. 5Embedding and distant relationship detectionExtract mean-pooled sequence embeddings using ESM2-650M via run_model to cluster related proteins or detect fold similarity below the pairwise sequence alignment twilight zone.

Start it here

Ask in plain language. The agent works on a Linux machine of its own — it installs and runs what the job needs, keeps every file, and reads the results back with the numbers explained. Nothing to set up, and a long run continues after you close the page.

What to watch for

• Ambiguity codes (X, B, Z, J) corrupt theoretical molecular weight, pI, and GRAVY; if sequence ambiguity exceeds roughly 2%, physicochemical properties are flagged as unreliable.
• Using the cystine extinction coefficient instead of the all-Cys-reduced value for cytoplasmic or DTT-reduced proteins leads to systematic errors in calculated concentration.
• Theoretical pI applies free-amino-acid pKa values to an unfolded chain, which can deviate by more than a full pH unit if residues are buried, modified, or fused to tags.
• Reporting percent identity without query coverage and aligned length is misleading, as high identity over very short segments indicates a shared motif rather than an orthologue.
• Pairwise identity between 20% and 35% sits in the twilight zone where folds may be conserved but functional equivalence cannot be safely assumed.
• Nucleotide inputs must not be evaluated with protein statistics; sequences containing 90% or more ACGTUN must first be translated using the longest stop-free open reading frame.

What it runs

Questions