Structure-aware sequence embedding
Create one 1,280-dimensional representation from an amino-acid sequence paired with 3Di tokens.
Start this methodRepresent proteins and compare structure-aware mutations with SaProt
Generate 1,280-value SaProt embeddings or score one exact substitution using amino-acid and precomputed 3Di tokens.
Choose an approach
Scientific approaches
Select the analysis that best matches your scientific question. Each approach opens with its relevant inputs and controls.
Zero-shot mutation preference
Compare mutant and wild-type log probabilities with the submitted residue masked while preserving its 3Di context.
Start this methodPrepare
Sequence with precomputed 3Di structure tokens
Use one annotated-sequence row containing equal-length amino-acid and Foldseek 3Di strings; mutation analysis also needs one XnY substitution.
- Exactly one row and at most 1,000 residues
- One accepted 3Di token per amino-acid residue; all-# structure-free input is rejected
- Mutation mode requires one valid one-based XnY substitution matching the wild type
Sequence column
sequence3Di column
structure_3diMutation column
mutation · for example V3AHow the analysis starts
Choose a Project, open a compatible Dataset, then select the rows you want to analyze. Ubi will open this method with the compatible controls and column mappings ready to review.
Configure the scientific method
These controls appear in the Dataset analysis panel, where values can be checked against the actual input before the run starts.
Analysis
Embedding or mutation preference
Choose the analysis that answers the current scientific question.
Column mapping
Sequence, 3Di, and optional mutation
Map the Dataset fields directly; this method does not generate or infer 3Di tokens.
Model context
SaProt 650M · deterministic inference
Use the same model and 3Di preparation when comparing candidates.
Review results as scientific outputs
Results open with the figures, structures, sequences, and metrics needed to answer the scientific question. Downloadable files remain available for downstream analysis.
How to interpret the result
- Embeddings are model representations for comparison or downstream analysis, not measured biological properties.
- A positive mutation log-probability ratio means the pinned model prefers the submitted mutant in that structural context; it does not establish improved function, affinity, stability, or safety.
- Compare results only when sequence preparation, 3Di generation, model bundle, and operation contract are identical.
1
Structure-aware embedding Dataset
Export the complete 1,280-value vector for clustering or downstream modeling.
2
Mutation-effect Dataset
Review the mutant/wild-type log-probability ratio, masked token, and exact token-block evidence.
3
Sequence and 3Di context
Review the amino-acid and structural-token context used for each result.
Continue your analysis
Use a complementary method on the same Project data.
Method scope and limitations
- SaProt requires supplied 3Di tokens and does not substitute amino-acid-only input.
- Embeddings and mutation preferences are model outputs, not measured biological properties.
