westlake-repl / SaProt

Saprot: Protein Language Model with Structural Alphabet (AA+3Di)
MIT License
359 stars 35 forks source link

Interpretation of predicted mutation effect #64

Open TangYiChing opened 2 months ago

TangYiChing commented 2 months ago

Hello,

Can you tell me how to interpret the score appropriately? Does it have a unit, like ΔΔG in kcal/mol? Is the score comparable? What is the difference between effects and probability scores?

LTEnjoy commented 1 month ago

Hi,

The model takes the "wild type sequence" and "mutation information" as input and outputs a "score". A positive score means the mutation is better than the wild type from evolution perspective.