BenchBB was initiated based on the insights gained from the Adaptyv Protein Design Competition to establish a benchmark for computational protein binder design.
It includes seven carefully selected targets: PD-L1, EGFR, IL7Ra, BHRF1, SpCas9, BBF-14, and MBP. Each target offers unique advantages for benchmarking, from ease of production, to structural challenges and diversity in binding modes.
For more details, read the paperBenchBB addresses the critical need for standardized benchmark targets in computational protein design, enabling consistent evaluation and direct comparison of different methods.
Targets are selected for novel interfaces, challenging conformations, therapeutic relevance, and ease of production.
Like benchmarks in machine learning (GLUE, SuperGLUE, MMLU), BenchBB provides a foundation for measuring progress and driving innovation in computational protein design.
Track and compare the performance of different protein design models across our benchmark targets.
Access comprehensive datasets, analysis tools, and visualization features to dive deep into the results.
Participate in regular challenges to test your models and contribute to advancing the field.
Use the Adaptyv Cloud Lab to benchmark your protein-binder designs against the BenchBB targets.
Choose which protein targets you want to benchmark against
Specify the number of designs or upload sequences for each target
Review your selections and proceed with your project
Choose which protein targets you want to benchmark against. Ideally, you should benchmark your model against all targets.