BenchStab

BenchStab queries and benchmarks web-based protein stability predictors through automated execution and result collection on lists of mutants.

Key Features:

• Automated Predictor Execution: BenchStab executes 19 web-based protein stability predictors and collects results automatically.
• Extensibility: The tool supports easy integration of additional predictors beyond the initial set.
• Independent Evaluation Dataset: BenchStab includes a dataset derived from FireProtDB for independent evaluation of predictors, intended to be excluded from future training and validation.

Scientific Applications:

• Protein Stability Prediction Benchmarking: BenchStab enables systematic comparison of multiple web-based protein stability prediction methods.
• Mutation Effect Analysis: BenchStab facilitates assessment of how mutations affect protein stability by executing predictors on lists of mutants.
• Independent Method Evaluation: BenchStab provides a held-out dataset for unbiased evaluation of 24 different prediction methods.

Methodology:

BenchStab was developed as a console tool and Python package. An independent dataset was created from FireProtDB for evaluation purposes.