Sequence Embedding

Sequence Embedding turns DNA, RNA, or protein sequences into numeric vectors.

Use it for

• comparing sequences by model representation;
• creating embedding tables for later analysis;
• testing local sequence models before building a larger pipeline;
• feeding embeddings into downstream custom tools.

Inputs

• FASTA/FA/FNA/FAA/TXT file, or an inline sequence.
• Molecule type: DNA, RNA, or protein.
• Max tokens.
• Compatible installed AI Model.

If both a file and inline sequence are provided, Liatir uses the file.

Compatible models

• Nucleotide Transformer v2 50M
• Nucleotide Transformer v2 500M
• ESM-2 8M Protein

Outputs

• Embeddings CSV.
• JSON summary.
• Sequence count.
• Embedding dimension.
• Mean sequence length.
• Provenance.

How to read the result

Do not try to interpret every embedding value manually. An embedding is a vector that becomes useful when compared, clustered, visualized, or passed to another tool.

Check:

• sequence count matches what you expected;
• embedding dimension is stable for a model;
• max tokens did not truncate important sequence context;
• molecule type matches the model.

Technical details

Tool ID: ai-sequence-embedding

Liatir pools model outputs with attention-mask mean pooling and writes a per-sequence embedding table. The output is a representation, not a final biological label.